Commons talk:Quality images candidates

From Wikimedia Commons, the free media repository
Jump to navigation Jump to search

Increase minimum resolution to 4K?[edit]

I think the 2MP resolution requirement is outdated. It was last updated in 2007 and it's roughly the resolution of a 1080p monitor. Today a 4K monitor (about 8MP) seems much cheaper than 1080p was then, the unclear rule results in pictures getting turned down based on the reviewer's arbitrary resolution requirements, and most importantly I think a quality image should fill up a modern computer screen. --Trougnouf (talk) 17:00, 4 January 2019 (UTC)

I have already wondered about it myself. It is obvious that with the current development of technology, the bar will have to be raised sooner or later. The only question is that is it comming a good time for this? And other question: what we do with pictures that are currently QI, and after lifting the bar will no longer have this distinction? Tournasol7 (talk) 21:07, 4 January 2019 (UTC)
imo QI's should always remain QI's, it indicates that a picture met our quality requirements at the time it was submitted.
One reason is that a quality image of a X taken ten years ago certainly won't look like what we expect with today's level of technology but X ten years ago is pretty much a different subject that can't be captured today.
Another reason is that QI's are the only timeless ones and that makes it much more motivating, sure we could make a category of images that no longer meet our requirements but our requirements are much more than the resolution and I don't see the community spending the resources necessary to nominate review and demote all these older images. --Trougnouf (talk) 21:31, 4 January 2019 (UTC)
per Commons:Image guidelines: "The purpose of quality image status is to recognize that at the moment of creation, a Commons user skillfully achieved a desirable level of quality, a recognition that is not erased by later advances." --Trougnouf (talk) 01:50, 5 January 2019 (UTC)
I'm generally in favor of raising the minimum resolution requirement and 4K sounds like a good yardstick. However, I would strongly suggest, instead of defining a minimum total resolution, to instead have a minimum length for each side of maybe 2000 px. This would avoid discussions for images in unusual aspect ratios. --MB-one (talk) 11:54, 5 January 2019 (UTC)
+1 for increasing the minimum resolution to 4 MP for all new QIC's starting with photographs taken from 2019 and later. ... But it was discussed very often without any result. --XRay talk 13:30, 5 January 2019 (UTC)
As XRay; increasing the minimum resolution to 4 MP for images taken from 2019 and later. Tournasol7 (talk) 15:20, 5 January 2019 (UTC)
I think 8MP is more appropriate since that roughly fills a 4K display but if the consensus is to start with 4MP then I'm all for it. --Trougnouf (talk) 13:40, 5 January 2019 (UTC)
  • Symbol support vote.svg Support even 8MP --A.Savin 16:21, 5 January 2019 (UTC)
  • Symbol support vote.svg Support for 4MP from 2019. Would even go for 8 MP since that might weed out many downsized photos and force them to upload at more or less original size. --Cart (talk) 16:42, 5 January 2019 (UTC)
  • Symbol support vote.svg Support Either 4MP, 1800x2000 or even 8MP. For now, I think we could all agree on 4MP. --Podzemnik (talk) 11:32, 10 January 2019 (UTC)
  • Symbol support vote.svg Support makes sense. 2MP is a wee bit too little. ― Gerifalte Del Sabana 10:31, 3 February 2019 (UTC)
  • Pictogram voting info.svg Info GerifalteDelSabana and everyone else: The real voting is a bit further down in this thread. --Cart (talk) 10:35, 3 February 2019 (UTC)
  • @W.carter: Haha, yeah... I figured. If you do scroll down you'll see that I actually voted there already. :) ― Gerifalte Del Sabana 11:55, 3 February 2019 (UTC)

Increase minimum resolution to 4K? - part 2[edit]

In May last year, I had a look at the stats for Featured Picture. (I haven't got figures for 2018 as I scraped the POTY list, which isn't out for 2018 yet, and I seem to have mislaid my spreadsheet. Here's what I found:

The above slide show demonstrates how the population of photographs at FP has changed over the years. The x-axis is the length of the shortest side. The y-axis is scaled to 100 for the most popular bar. The red line shows the 5th percentile. The grey bars are for images with shortest side less than 2000 px. If you wrap-around the slideshow to flip between 2007 and 2017, the difference is dramatic.

In 2006 the Featured Picture / Quality Image criteria specified that images should be at least 2 megapixels. Your television used a cathode ray tube but you wanted a flat screen. The resolution of your television was 720 × 525 (0.38 MP). High definition TV, at 1920 × 1080 (2.1 MP) was a few years away yet. Today your television uses an LCD and if you bought one recently it will be 4K, at 3840 × 2160 (8.3 MP). This is about 4 times the linear resolution of standard definition. There are a few sizes of desktop and laptop display resolutions, and the market is led by 4K models. A 2 MP image will only occupy one quarter of the display on a 4K display and is only good to print a small 15 × 10 cm photo.

If we choose a higher threshold, there are two options of measurement. One is to use megapixels as before, and another to use the length of the shortest side. While megapixels are easy to understand, they don't scale linearly, and the file description page does not display MP as standard (though there is a script for that). Linear resolution is a better measure of how much an image can be magnified before it degrades, and the shortest side is the limiting factor. That value is also easy to discover from the file description page. English Wikipedia Featured Pictures set a minimum of 1500 px on the shorter side six years ago in 2012.

A 3:2 ratio six megapixel image has a shortest side of 2000 px. For five megapixels, the value is 1826, which we could round down to 1800. A threshold of 2000 px on the shortest side would affect only 4.5% of 2017's featured images. A threshold of 1800 px affects only 2.1% of 2017's featured images. Most, but not all, of these images have been downsized-for-the-web. Some come from Wiki Loves Earth, which didn't ask for high resolution, and some come from external sites where downsizing is routine.

The Commons FP rule says "Images (with the exception of animations, videos, and SVGs) of lower resolution than 2 million pixels (pixels, not bytes) are typically rejected unless there are strong mitigating reasons". The allowance to break the threshold for some images/circumstances can be retained, and argued for by the nominator. QI doesn't have that exception.

I think choosing an 1800 px or 2000 px threshold on the shortest side would be a modest improvement on our current 12-year-old standard. A landscape image with shortest side 2160 will vertically fill a 4K display, so even a 2000 px threshold is not quite state-of-the-art for displays. An increased threshold will hopefully encourage less downsizing, and less discussion per-image about what is an acceptable bare minimum. It does not, of course, prevent any Featured Picture reviewer objecting about image sizes that exceed this, since even this threshold often represents a considerable downsize or crop. -- Colin (talk) 17:19, 5 January 2019 (UTC)

I think this suggestion is reasonable, so I mildly support it. One positive aspect of it is that at this point, there is jury nullification of the prohibition on downsizing, depending seemingly on who is doing the downsizing. That would be a bit less of an issue with this change, but nevertheless, I would propose that, along with this change, we would do away with a prohibition on downsizing at QIC, since in practice, there is no such prohibition, and the ostensible prohibition shouldn't be usable as a cudgel against some users, with others being given license to do it to their hearts' content. -- Ikan Kekek (talk) 03:40, 6 January 2019 (UTC)
Ikan I don't know about the practice of doing, ignoring or rejecting downsizing at QIC. The reason for discouraging downsizing is that strong downsizing impacts on current and especially future uses for the work. As we can see from the chart, and the technology changes, the demand for higher resolutions isn't stopping. See User:Colin/PixelPeeping for some sizes need for certain print or display purposes. Even a 4K monitor requires a 3840 × 2560 (9.8MP) 3:2 image, which could then be cropped to 16:9. If you look at the charts above, you see that 2000px is a very common size. So if we start seeing more people downsize to 3000 × 2000 precisely because it meets QI minimum and we no longer discourage downsizing, then we've shot ourselves in the foot -- today's QI images won't even fit today's TVs. And a 300dpi print needs 2475 × 3713 (9.2MP) for an glossy magazine page, or 5007 × 3338 (16.7MP) for a double spread. So plenty good reasons to encourage people to upload large. The other issue with downsizing is that people do it because reviewers pixel-peep. Where I do think some downsizing is reasonable is where the image had inevitable bad conditions (very high ISO, very distant -- so lots of atmospheric shimmer) or where it is stitched and stretched. Perhaps an idea is for QI to have a link to the image at 2000px on the shortest side, and for all reviews to be conducted at that resolution. The most common size for DSLR images is currently 6000 × 4000 (24MP) which means many images would be reviewed at about 50% size, which nicely hides many pixel defects while still being large enough to judge focus, sharpness, etc. -- Colin (talk) 11:39, 6 January 2019 (UTC)
The problem with a shortest-side rule is that for example, a 2000 × 2000 technically passes while a 5400 × 1800 (think someone getting a panorama straight out of camera by cropping a 19 MP image) would not. Looking good on a 4K display should not be our aim; super-panoramas, even something like 40000 × 5000, would not look good on a 4K display. I think a good standard would be "no part of the image is unsharp at the pixel level (unless intentional) when scaled to 4 MP." -- King of ♠ 15:38, 6 January 2019 (UTC)
King of Hearts I don't see how you can complain about 2000 × 2000 being an issue (I don't think it is) yet think 4MP is OK. The big problem with MP is that, unless you add some JavaScript, the file description doesn't say the size in MP and unless you have a spreadsheet to hand, and know the arcane syntax of the MediaWiki thumbnailer URL format, it isn't easy to work out how to scale any image to 4MP. That, to me, makes any rule that isn't directly obvious a non-starter. And, to be honest, anyone creating a horizontal panorama, using one row of landscape-oriented frames, needs to learn something about technique. So I'm not really sympathetic to the 1800px height issue. -- Colin (talk) 20:47, 6 January 2019 (UTC)
In practice, a lot of pixels are "wasted" in a Bayer-interpolated image. So a 100% perfect 4 MP image has about as much detail as a 6 MP or even 8 MP image judged with the typical allowances at QIC. The main problem we have right now is that typical allowance doesn't seem to scale up, so many people are judging a 12 MP image and a 36 MP with equal rigor at the pixel level. -- King of ♠ 21:17, 6 January 2019 (UTC)
I think calculating the MP count as weight x height is a non-issue, the vast majority of the time it doesn't come into the equation (much bigger images) and when in doubt then it's very simple math. Now judging different formats on an equal playing field is a different matter, but I don't think that's ground for penalizing unusual formats. In my opinion File:Centrum De Calidris in Middelkerke, Belgium (DSCF9885).jpg is as qualified as any 4MP image even though it's 1005 pixels tall (@Colin this resolution doesn't just occur from a lack of technique, it's not always necessary to stitch a panorama when a relatively small 16MP sensor can be cropped to 1/4 band and still output the desired level of detail), same with File:Villa Doris, Middelkerke (DSCF9903).jpg's 1595 pixels width yet 7 MP of useful information. Diversity leads to great things and it's worth the extra effort to sometimes open the calculator or judge an image differently imo. --Trougnouf (talk) 21:45, 6 January 2019 (UTC)
Trougnouf, any maths is too much maths. Unless the UI explicitly tells users "this is too small" then we waste our time -- either QIs wrongly promoted or FPs voted on before someone objects. The problem with any threshold is that someone will point at this image or that image and say "that's a useful image" or "that's a quality image". You've got to set it somewhere and that will mean a few nice pics are below the threshold. The stats show that images with < 2000px on the shortest side are below the 5th percentile of FP promotions in 2017, and likely even lower for 2018 never mind 2019. A MP threshold is not only harder to compute but illogical when you consider that all that matters for quality is linear resolution. A 1000px tall panorama can't be printed large no matter whether the width is 2000px or 10000px. Let's leave MP for camera manufacturers to boast with, because grows quicker than linear resolution. -- Colin (talk) 22:32, 6 January 2019 (UTC)
On downsizing: I understand the argument against it, but since actual practice at QIC is to accept it unless people feel like not accepting it, it's a waste of time to have a rule that's routinely flouted. Ikan Kekek (talk) 01:43, 7 January 2019 (UTC)
  • Symbol oppose vote.svg Oppose For wildlife images, a blanket 4K is inappropriate. Cropping improves wildlife images and make them much more useful, particularly on mobile devices. Also, cropped images are essential to illustrate Wikipedia articles and for use in other Wikipedia projects. Many wildlife images (esp. macro) are also rotated and that inevitably demands cropping. Cropping is essential for VI images. Some users are having wildlife images promoted at QI by not cropping them so they can meet today's size threshold.
    My images are used in presentations on large screens and they do not all fill the screen on their own. One screen can have several images and text. Likewise, where my images are used in publications, they are seldom two-page spreads. If they are, they would be 4K images anyway, not animal portraits.
    I am sure 4K is sensible for huge numbers of QICs and I would support a policy that nominators should declare their images have not been downsized.
    A more useful modification to QI might be to raise the 'composition quality' threshold, but that's another debate.
    And finally. Pixel-peeping. In assessing an image (especially FP) it is important to zoom in to assess what level of post-processing has been employed. In wildlife images for instance, aggressive blurring of the background, even false backgrounds, are presented without disclaimers. Assessing a focus-stacked or stitched image also requires very detailed examination. Charles (talk) 11:09, 10 January 2019 (UTC)
Charles, there are several options to consider, not just 4k. Commons is a media repository so if you think a particular use of an image would be better presented cropped, then that's not strictly Commons concern. We can host multiple versions of any file, and wikipedia articles can use a cropped alternative, and there is even a template on Wikipedia to display a cropped portion of an image. Many professional photographers retain (or offer as stock) uncropped photographs because they permit a publisher to use the image landscape, portrait, or to overlay caption text or even titles onto the image. When all one offers in a repository is a tight crop, you have made choices a publisher may not agree with or find useful. I looked back through dozens of your FPCs and can only find a few examples where the shortest side was < 2000 px. All of those could be larger if a slightly bigger crop was used. Indeed some of those had very short height and wide width, which is exactly the wrong kind of cropping for mobile devices and for Wikipedia, which favours a portrait orientation and finds wide landscape to be problematic. The purpose of QIC or FPC isn't to have rules that permit ALL possible good or great images to be promoted, but to maximise the good stuff. Currently our 2MP limit encourages downsizing to an extreme that is frankly no longer useful in 2019. I think a "2k on the shortest side" rule would limit what you can nominate only marginally (a few %) and if it encourages you to not crop vertically so tightly, I'd count that as a bonus. -- Colin (talk) 13:00, 10 January 2019 (UTC)
I crop tightly for online use. Where my images are requested for print media, I am often asked for the RAW file or alternatively I supply a CMYK version (often uncropped) as all my uploaded images are sRGB. Charles (talk) 14:21, 10 January 2019 (UTC)
That only proves my argument and you are not doing the online users any favour by restricting their choices. I'm not saying we shouldn't crop for artistic reasons, but a tight crop can be restrictive, and your argument that we should have a low minimum for FP/QI simply so you can crop tightly isn't a good one. I also looked through dozens and dozens of JJ Harrison's recent uploads and only found a few that were just below the "2000px on shortest side" suggested limit. So I don't think in practice it would be any problem for either wildlife photographer. And at FP we have always had the "mitigating factors" to let reviewers consider images below any threshold. -- Colin (talk) 15:07, 10 January 2019 (UTC)
  • Symbol support vote.svg Support Yes, please, let's increase the minimum MPx requirement once for all --Poco2 18:09, 10 January 2019 (UTC)
  • Symbol support vote.svg Support Yes, please. 2MP was huge back in the days that rule was created. Nowadays 2MP is tiny compared to the sensors available and the screens being used. I can crop my sports photos and if they have too much noise I can still downscale them to 3000 pixels on the long side which results in 6MP of total resolution. I'm totally fine with 4MP and could even arrange with 8MP since most of my sports photos made indoor and downscaled won't yet be good enough for QIC anyways. --Granada (talk) 18:36, 10 January 2019 (UTC)
  • Proposal: 4 MP in 2019, 6 MP in 2020, and 8 MP in 2021 (where it will remain for the foreseeable future). This will allow a smooth transition to get people accustomed to the new standards; Rome wasn't built in a day. I think 8 MP is a good final goal for most pictures, though keeping 4 MP for moving subjects may be adequate (if we do decide to go with multiple standards, we'll have to define what "moving subject" means and think of any other exceptions to carve out). Meanwhile our standards for images with more pixels will also increase; right now if a D850 photo looks not so great at 100% but perfect at 2 MP I have no real reason to say no to it, but the standards for true resolution will rise with this change. -- King of ♠ 01:32, 11 January 2019 (UTC)
  • Symbol support vote.svg Support 4K, but Symbol oppose vote.svg Oppose 8K. We can always have another discussion in 2 years. Regards, Yann (talk) 05:37, 11 January 2019 (UTC)

Increase minimum resolution to 4K? - part 3[edit]

  • Pictogram voting comment.svg Comment Could we all please stop voting. I don't see any consensus for one proposal. Is it 4k, 4MP or 2000px on shortest side, 8k or something else. Voting won't solve that. 4k is a 16:9 landscape-format standard for TV. What does it mean to "fill up a modern computer screen"?
  • If a 3:2 landscape image fills a 4K screen horizontally, with some cropping, then it needs to be 3840 × 2560, which is 10MP.
  • If a 3:2 landscape image fills a 4K screen vertically, with black bars, it needs to be 3240 × 2160, which is 7MP.
  • A 3:2 portrait that fills a 4k screen vertically would be 1427 × 2140, which is 3MP.
  • A 3:2 portrait that fills a 4k screen horizontally, with cropping, would need to be 3840 × 5760, which is 22MP.
I don't think "4k" is a reasonable threshold for judging photos that come in all sorts of shapes and sizes. It is just too variable and arguably "filling the screen" only makes sense for an image close to 16:9 landscape. It is fine if someone wants to make 4K the minimum standard for video at FP.
Another common usage is mobile devices, and for that a vertical format is preferred. A Samsung Galaxy S8 is 2960 × 1440, the largest iPhone is 2688 × 1242. And, well, the pixels are so small that all our pixel peeping nonsense becomes truly ridiculous.
I still think MP is hard to implement as a threshold for anyone to policy, and suspect most people have been assuming most nominated images are well above the threshold (so don't care to do the maths unless resolution looks suspiciously low) or that someone else at FPC will check for them (so don't care to do the maths). If we increase a MP threshold then it becomes more likely you'll have to do the maths. And as I've said above, I don't think MP has any bearing on quality of ability to use an image for a purpose. The high-quality printing standard is 300ppi, which is a linear resolution for a given page size and format.
Anything we pick will be a compromise in some way. I think it should serve as a nudge to improve standards a bit and be trivially easy to police. I don't think a 4MP standard will make an ounce of difference because there are so few nominated pictures that small currently. The charts above show that in 2017 a 2000px on shortest side was exceeded by 95% of images at FP. -- Colin (talk) 10:15, 11 January 2019 (UTC)
Obviously I mean 4 MP. I don't see your argument about "doing the maths" useful here. 1. There is gadget which could be made on by default, 2. It allows some unusual format like File:Line scan photo of Shinkansen N700A Series Set G13 in 2017.png and File:Line scan photo of nine car BART C1 train in 2017.jpg. Regards, Yann (talk) 10:49, 11 January 2019 (UTC)
Yann, I'm not aware of a gadget for MP, only some JavaScript -- which gadget do you refer to? If you get a MP gadget on by default for all users or the File Description page amended to show MP then that removes the "it is hard to work out" argument, but neither of those are remotely likely any time soon. In the mean time, that argument holds: few of us can do that sort of maths in our heads. And it still doesn't change the fact that MP is a very poor proxy for quality. The two examples you give are quite exceptional (being 58:1 aspect ratio), which FP permits for such an unusual photographic device (and really, no reason for QI not to permit exceptions). A 4MP rule permits me to crop my 24MP 3:2 photo to be just 667px tall to create a "panorama" that QI would have to accept. A 667px tall image is of very limited usefulness and only vertically "fills" a third of a 4K monitor. -- Colin (talk) 12:35, 11 January 2019 (UTC)
I believe there’s a fundamental disagreement here about how much the longer side should matter when measuring resolution. Everyone agrees that it should matter less than the shorter side. The MP crowd (myself included) believe that the amount the longer side should matter is proportional to the length of the shorter side (so on a 3000 x 2000 image, lengthening the longer side by 1 pixel gives an extra 2 kP while lengthening the shorter side by 1 pixel gives an extra 3 kP). However those advocating for a minimum length for the shorter side are saying that the longer side should not matter at all. I personally feel this is too extreme. -- King of ♠ 17:25, 11 January 2019 (UTC)
King of Hearts what changes the quality, at a pixel level? What makes one image sharper than another or more detailed? What makes the noise or CA troublesome or irrelevant? I don't make an image softer by cropping more or sharper by cropping less. Changing the shape of an image doesn't change the technical quality at all. You talk of "lengthening" one side, as though one could lengthen an image in one dimension. Resizing (downsizing, upsizing) changes both dimensions together. You can only shorten or lengthen one side by cropping more or less, and then it becomes a different image, with a bit more or less content, but the actual content doesn't change in quality one bit. Linear resolution is the limiting factor on how big you can make an image. A 1000px tall image will only reach halfway up a 4K monitor, or print highest quality only 8.5cm tall, no matter how wide it is. I think shortest-side balances better the needs of panorama photographers (wider than 3:1) vs wildlife photographers (who create images between 3:2 and 1:1). The latter really get stuffed by a MP rule because they are technically challenged and don't have aspect-ratio multiplication giving them any boost. The former get a huge boost by wide aspect ratio but really aren't technically challenged at all -- it is trivially easy to create high resolution panoramas (specialist slit cameras excepted). And for all the others, churning out 3:2 or 16:9 images, we still have an easy-to-apply rule that will scale linearly as cameras and displays offer and demand more resolution. -- Colin (talk) 18:31, 11 January 2019 (UTC)
That is actually a good point about it being a convenient standard for distinguishing between wildlife and landscape, though I still disagree with your rationale — for me, megapixels = content/detail and an image needs to have a minimum amount of content to pass QI, rather than being fixated on what can be viewed on a particular type of monitor. Anyways let's just say we've adopted a 2000px shorter side minimum. There are still two points I'd like to make: 1) Add an alternative standard of 8 MP where an image only needs to meet one or the other to qualify, so that no image will need to have more than double the total amount of detail than a "most favored image" (1:1) to pass QI. 2) These days there is a perverse incentive for people to downsample their images to look good at a smaller resolution. I almost never do it myself for philosophical reasons, but I agree with them: if it looks perfect at 2 MP, then I pass the image. "You could have done better" isn't really a valid objection at QIC, and if it meets the minimum requirements with no flaws at that size then it should be enough (given whatever the current standards are). Sadly most people would fail a 36 MP image that looks like it has only 9 MP of detail (50% of total possible linear resolution). Under a higher resolution requirement, there may be the opposite perverse incentive: If you took an image that required heavy cropping to say, 1800 x 2700, but is tack sharp at that resolution, you are incentivized to either crop less, making the composition slightly worse (QI doesn't care about composition as long as it doesn't suck), or to upsample it to 2000 x 3000 (if you don't believe me, take one of your own pixel-sharp images and upsample it by 10% linearly; I bet you can barely find any unsharpness).
Perverse incentives have always been my pet peeve at QIC; even if no one attempts to game the rules, it's still very unsettling for an image to not pass when a worse or basically unchanged version of the same image would pass. I care much less about whether standards are high or low than about those standards being applied fairly to each image. And I mean this in two senses: the standards need to be fair, and they need to be objective enough that an image's fate does not (primarily) rest on luck of the draw based on who reviews it. Subjectivity is impossible to stamp out of course, but we should at least try. (By the way, I commend the effort to raise standards, which helps mitigate "pixel-peeping reciprocity failure," where people do not judge landscapes with half the linear resolution twice as harshly.) -- King of ♠ 02:39, 12 January 2019 (UTC)
King of Hearts, I prefer to judge content by eye, rather than assume counting total pixels means anything useful -- quite a lot of it could be sky. I guess the two main drivers for a standard are to ensure the image has resolution enough to be widely useful and to discourage significant downsizing. If all our images were 3:2 then there wouldn't be a difference between MP and shortest-side other than convenience. So for me it is important that the rule doesn't punish those who have technical difficulty meeting the standard and doesn't reward those with no difficulty for either downsizing or using inferior technique (e.g. creating a panorama by just cropping a single frame -- wouldn't impress many folk at FP). I think your idea of having two measures is a good one to consider. This is also where QI an FP might diverge: I'm still struggling to consider that an image in the bottom 5% of height, and only going halfway up a modern PC monitor, could be considered "the finest on Commons" no matter how wide it might be. If you look at the early days of FPC, there were some tiny FPs with lovely pictures that I'm sure some people thought were great, but they are really pathetic now. A bit like awarding a prize to a postcard when what you want is a print to hang in a gallery that is several feet across.
I don't know how to solve your concern about gaming the system, and don't see that MP or shortest-side makes a difference to whether one can tweak an image to get it to pass. Perhaps worry less about images that are on the threshold whether they pass or not. They are a small percentile. As long as most of the images passing are good quality and most of the images failing are bad quality. -- Colin (talk) 15:46, 12 January 2019 (UTC)
The two cameras I use a Nikon D3300 and D3400 both 24.2mp. I've gone through my QI images most are over 12mb with the exception of 1 around 8mp. There should a dicussion on what camera is best used. Because everyone has there favourite camera or is unable to buy new equipment. This would weed out any devices such as iPhones, Samsung phones etc. Adamdaley 08:04, 12 January 2019 (UTC)
— Preceding unsigned comment added by Adamdaley (talk • contribs) 08:04, 12 January 2019 (UTC)
I don't think the camera used is important, and in good light, a modern smartphone with quality camera should be capable of producing a QI image. The problem with only looking at one's own portfolio is that you don't then appreciate some of the problems faced by people who take different images. Wildlife requires very expensive telephoto lenses and even then some cropping, whereas landscape or architecture can be done with cheap equipment and the full frame. Low light or high speed events such as sports or celebrity shows require high iso and some forgiveness of sharpness that wouldn't be considered acceptable for a studio portrait, say. -- Colin (talk) 15:46, 12 January 2019 (UTC)
  • Nobody who is a common user of an Iphone etc, will take the time to photoshop their images. I know I wouldn't, because it's basically a phone. Will the entry of a DSLR Nikon Range improve? For my cameras, they are both 24.2mp. The Nikon D3500 is 24.2mp and i don't see by 2021, it'll have changed much since my D3300 is the only cameraican use for GPS coordinates. Adamdaley 02:53, 13 January 2019 (UTC)
    — Preceding unsigned comment added by Adamdaley (talk • contribs) 02:53, 13 January 2019 (UTC)
  • Adam, I'm not really sure what your point is, but also I don't think it is relevant to this discussion. Photoshop Lightroom is available for mobile devices and the desktop Lightroom now has a cloud version because mobile photography is a huge thing. Actually, a lot of iphone users never make calls. It's a $1000 camera/computer that fits in your pocket. -- Colin (talk) 11:14, 13 January 2019 (UTC)

Discussion[edit]

Just to clarify, this voting is just for QI? Because it's probably a lot less practical for FP, given that it may include things that are scans and the like, and there may not be any significant actual benefit of more resolution for scans of smaller artworks past a certain point, other than making them next to impossible to get. Adam Cuerden (talk) 04:30, 11 January 2019 (UTC)

It is a bit of both. What is a common resolution/size of scans you get? -- Colin (talk) 10:15, 11 January 2019 (UTC)
It varies a bit, but a lot of it's bound by the practices of the archives that host them, and we will be cutting out a lot of possibilities by raising the bar too far. Looking just at my most recent few FPs here, giving the year before each:
Due to the nature of the photographic grain on these, I see little value in any higher resolution. And, yes, some of these are definitely better than others, but we exist to serve Wikipedia, so subjects matter.
Just to give a few more examples, looking more at variety this time.
I think the point's made. In historical imagery, archives' practice just give the pixels called for, nor would adding pixels add meaningful detail in many cases. That said, most of these are at least close.
Also, from a procedural ground, I do think if it's meant to stick for FPs, that should be a separate vote and would need to be done with notifications at the top of Commons:Featured picture candidates and the other-language versions. This was never framed as necessarily applying to FPs, after all. However, I do think the suggestion is a reasonably good idea for QI, as it focuses almost exclusively on contemporary images taken by our photographers. But, it just simply doesn't have the same width of scope featured pictures does. Adam Cuerden (talk) 08:50, 12 January 2019 (UTC)
@Adam Cuerden: Agreed, FP is a different matter. This proposed change is for QI only. --MB-one (talk) 12:27, 12 January 2019 (UTC)
Yes, what I meant was the discussion was a bit of both. I don't think we are at a point where a clear proposal is being voted on (despite the vote icons above, they are all for different things). I agree there should have to be a vote at the FPC for those rules, though it would be convenient if they are similar, there is no reason why they need to be. I've never really understood why your restorations couldn't be considered for QI -- the whole project doesn't make a whole lot of sense to me TBH. Wrt some of the scans being below 2000px or 4MP, I agree there is not much point in scanning much higher than the grain resolution of the original. But then, not every image needs to be an FP and we do have many scans at very high resolution. So perhaps it is just "tough" that a nice picture is too low resolution -- much like us photographers have to accept that the weather was rubbish when we visited somewhere, or it was covered in scaffolding or had too many tourists -- not much we can do about it. -- Colin (talk) 15:24, 12 January 2019 (UTC)

Resolution of the discussion[edit]

Since the discussion has stalled, I think it is time to vote. I guess, it would be best to vote for or against both a minimum resolution (Option 1) and a minimum pixel length for the shorter side (Option 2). --MB-one (talk) 13:02, 26 January 2019 (UTC)

Voting period should be 14 days from now and end on Feb, 9th 2019 23:59 (UTC).

Option 1: Minimum resolution[edit]

“Newly promoted Quality Images should have at least 4 real megapixels of information (with the exception of animations, videos, and SVGs), for example, 2500 × 1600.”

Support
  1. --MB-one (talk) 13:02, 26 January 2019 (UTC)
  2. --Trougnouf (talk) 16:50, 26 January 2019 (UTC)
  3. A good compromise for 2019, we can explore higher resolutions in the future. -- King of ♠ 07:05, 27 January 2019 (UTC)
  4. --XRay talk 17:52, 27 January 2019 (UTC)
  5. --Tournasol7 (talk) 21:41, 27 January 2019 (UTC)
  6. Yann (talk) 09:11, 28 January 2019 (UTC)
  7. Though it's still not enough. --A.Savin 09:45, 28 January 2019 (UTC)
  8. Better than 2 Mpx -- Basile Morin (talk) 08:05, 1 February 2019 (UTC)
  9. Gerifalte Del Sabana 10:32, 3 February 2019 (UTC)
  10. I'd prefer Option 2 - but either way I'd like to see commons get over these 2MP. --Granada (talk)
  11. A small step in the right direction Poco2 14:08, 16 February 2019 (UTC)
Oppose
  1. Because the commons UI does not report megapixels, making this tedious to calculate, and because 4MP is still way too small for a 1:4 panorama. We got away with MP when the threshold was very low, but now it will be necessary to get out the calculator on more images. -- Colin (talk) 16:38, 26 January 2019 (UTC)
  2. Not self-evident with this interface and Mpx is constantly confused with MB, so I would be happy for another way of expressing minimum resolution. But I'd rather have this than option 3. --Cart (talk) 17:16, 26 January 2019 (UTC)
  3. Per Colin. --Smial (talk) 14:07, 28 January 2019 (UTC)
  4. Fine for inanimate objects, but inappropriate for wildlife e.g. for this type of image. Charles (talk) 11:45, 30 January 2019 (UTC)
  5. The picture is important, not the count of the pixels. --Stepro (talk) 11:47, 30 January 2019
  6. For the reasons others have given. - Jmabel ! talk 16:36, 30 January 2019 (UTC)
Abstain

Option 2: Minimum pixel length[edit]

“Newly promoted Quality Images should have a minimum native length of 2000 pixels on the shorter side (with the exception of animations, videos, and SVGs), for example, 3000 × 2000.”

Support
  1. --MB-one (talk) 13:02, 26 January 2019 (UTC)
  2. Shortest side is easy to determine and this method is kinder to wildlife photography that struggles, and harder on panorama photography that has no difficulty exceeding it with the right technique. -- Colin (talk) 16:38, 26 January 2019 (UTC)
  3. A clear way of defining min res. Modern panoramas shouldn't have any problems with this either. --Cart (talk) 17:18, 26 January 2019 (UTC)
  4. Per Colin, though I would have preferred 1.800px, which would allow slightly cropped images from 6-Mpix-DSLR e.g. in case of necessary perspective correction. Many of these cameras are still in use.--Smial (talk) 14:13, 28 January 2019 (UTC)
  5. Easy. We don't need calculator, just to read. Very soon, better cameras will come with more pixels -- Basile Morin (talk) 08:07, 1 February 2019 (UTC)
  6. Per 3000 pixels on the long side which is what I prefer for downscaling my images during export in Lightroom. :) --Granada (talk) 08:45, 10 February 2019 (UTC)
  7. A small step in the right direction Poco2 14:09, 16 February 2019 (UTC)
Oppose
  1. I simply do not agree that a 12.96 MP 1800 × 7200 panorama should fail the QIC criteria. -- King of ♠ 07:05, 27 January 2019 (UTC)
  2. Fine for inanimate objects, but inappropriate for wildlife e.g. for this type of image. Charles (talk) 11:46, 30 January 2019 (UTC)
  3. Long small objects are existing. --Stepro (talk) 11:48, 30 January 2019 (UTC)
  4. Oppose, there can be great long but not high panoramas. - Jmabel ! talk 16:37, 30 January 2019 (UTC)
  5. per Stepro. ― Gerifalte Del Sabana 10:32, 3 February 2019 (UTC)
  6. Initially abstained but I have yet another example of single shot "panorama" that is this time 8 MP yet below 2000 pixels high, File:Arcade du Cinquantenaire (DSCF7405).jpg, it's definitely at least a QI and I think it would be ridiculous to force Users to stitch multiple pictures panoramas (long process which introduces perspective distortions and stitching errors) or to uncrop the sky / road / unsightly bits for the sake of not multiplying H by W. --Trougnouf (talk) 08:48, 10 February 2019 (UTC)
Abstain
  1. I prefer the MP count but definitely not opposed to this--Trougnouf (talk) 16:50, 26 January 2019 (UTC)

Option 3: Minimum pixel length or minimum resolution[edit]

“Newly promoted Quality Images should have a minimum native length of 2000 pixels on the shorter side (with the exception of animations, videos, and SVGs), for example, 3000 × 2000. Alternatively an image may meet the minimum resolution requirements by having at least 8 real megapixels of information, for example, 1800 × 4500.”

Support
  1. As discussed below, I think this is a reasonable compromise that ensures a 4 MP minimum for all images with higher requirements for wide landscapes, while ensuring that images with twice the information as the minimum can still pass. -- King of ♠ 21:22, 27 January 2019 (UTC)
  2. A small step in the right direction Poco2 14:09, 16 February 2019 (UTC)
Oppose
  1. Too complicated. Rules like this should be easy for anyone to understand. Try thinking about a non-English speaking newbie who speaks a language the COM:QIC, COM:IG, etc. is not yet translated into (there are many...). --Cart (talk) 21:34, 27 January 2019 (UTC)
  2. have to side with Cart here. Let's not make it too complicated. --MB-one (talk) 08:59, 28 January 2019 (UTC)
  3. Per W.Carter. --Smial (talk) 14:15, 28 January 2019 (UTC)
  4. Fine for inanimate objects, but inappropriate for wildlife e.g. for this type of image. Charles (talk) 11:46, 30 January 2019 (UTC)
Abstain

Comments[edit]

I'm not clear what "real megapixels" are or what "native length" is. Is this referring to upscaled images that weren't "really" that size? If so, I think it would be hard to prove an edge case, and a grossly upscaled image will be soft. So could we just simplify the criteria to drop "real" and "native"? -- Colin (talk) 16:38, 26 January 2019 (UTC)

I think they are trying to convey a point which I've always supported - judge an image by how much detail it actually resolves, rather than resolution or sharpness alone. For me an image is QI if it looks perfect at minimum resolution, regardless of how soft it is at 100%. -- King of ♠ 07:11, 27 January 2019 (UTC)
@Colin: I took the "real" part from the current image guidelines. To me, it means exactly, what King of Hearts described. "Native" means "as detected by the image sensor. Both wordings are meant to ban blow-ups of pixels. --MB-one (talk) 10:47, 27 January 2019 (UTC)
Technically, a Bayer sensor does not image anything like the resolution of the resulting JPG , with the colour information resolved at a much lower frequency and luminance guessed from each colour. Additionally, a JPG does not store colour information at full resolution. Well, I am rather surprised if everyone has been judging QIs at 2MP, giving the widespread pixel peeping rejections on 24MP images. -- Colin (talk) 17:42, 27 January 2019 (UTC)

Wrt panoramas, one has to set a threshold somewhere. There will be images some regard as useful and good enough quality below that. A 4K screen has 2160 vertical pixels so a 1800 × 7200 will not come close to vertically fill a modern TV screen or computer monitor. There is, today, absolutely no reason to offer a panorama at only 1800 × 7200. A 24MP DSLR takes 4000 × 6000 images, so even allowing for some cropping and non-optimal technique of using landscape frames, should easily allow panoramas of 3000 px tall. Even an ancient 6MP camera can produce better panoramas if one uses portrait frames. This is 2019 so lets now set standards that could have been exceeded 14 years ago: File:Sydney Harbour Bridge night.jpg by User:Diliff in 2005. -- Colin (talk) 17:42, 27 January 2019 (UTC)

Again, I don't see why we should take a 4K screen as the standard when it is totally inappropriate for displaying ultrawide panoramas. Even a gigapixel 4:1 panorama will show up as only 960 × 3840 resolution (3.7 MP) on a 4K screen. I just want to set consistent standards: I'm fine with counting out 1800 × 7200, but 2000 × 2000 will have to go as well. What might work is something like "2000px or 8 MP" or "1800px and 6MP" (numbers are just an example). But before that, consistent standards is also why I want to go slower; an image nominated in March 2019 isn't suddenly much less worthy of QI than an image nominated in January 2019. 4 MP, a doubling of square resolution, is a good first step in my opinion. -- King of ♠ 19:30, 27 January 2019 (UTC)
If you display an ultrawide panorama such that it fits the screen horizontally, then the vertical height is a consequence of aspect ratio, not resolution. So that's beside the point. I'm talking about someone trying to look in detail at the panorama and panning along it horizontally. A 4MP limit permits a 1000×4000 panorama that doesn't even reach halfway up the screen and can only be printed a few inches tall. We exceeded that level by some margin 14 years ago in 2005, and the 2MP limit has hung around long past its merit, so we certainly aren't rushing this. I don't think taking baby steps or worrying about rules one day and then the next is relevant. We should agree what is high quality today. And if a 13742 × 3431 panorama was achievable in 2005 then there is no excuse for a <2000px panorama in 2019. Whereas today it is wildlife photographers who tend to create close-to-square aspect ratios and it is them who really do struggle to get 4MP, even with $12,000 lenses. I would consider '"2000px or 8 MP"' as a compromise at QI, though I really do hope not to see any 1000 × 8000 letterboxes. For FP, I don't see any reason to compromise for panoramas: we've expected very high resolution panoramas for some time now. -- Colin (talk) 19:45, 27 January 2019 (UTC)
Agreed on FP, but there because you can oppose for any reason whatsoever, we already have dynamic standards de facto, so I see no need to change the official rules de jure and cause wildlife photographers stress over whether to crop to 2000 × 2500 to meet the letter of the law or to crop to 1800 × 2250 to achieve the composition they really wanted and let voters decide whether that's a resolution they are willing to accept. -- King of ♠ 21:14, 27 January 2019 (UTC)
Pictogram-voting-question.svg Question You are debating 1800 vs 2000 on the short side of a panorama. That's just 200px difference. Roughly, how common is it for users to upload panoramas at min height of 1800px or thereabout? (As in today and back in 2018, not historically) Is the larger part of uploaded panos on this crucial limit or are there just a few? --Cart (talk) 21:43, 27 January 2019 (UTC)
Actually, I think a reasonable standard would be: Forget the nominal resolution or dimensions of the image. Just downsample or upsample every photo to 2000px on the short side when reviewing, and it passes if it is sharp in all areas that need to be sharp. That way, if a 1800px image (4.86 MP at 3:2) genuinely has perfect pixel-level sharpness, then it can tolerate a 10% linear upsampling and will pass, while a 2400px image which is a bit fuzzy (8.64 MP at 3:2) will fail; I'm talking about a level of fuzziness which would be acceptable at the pixel level on a 4000px image (24 MP at 3:2). -- King of ♠ 23:27, 27 January 2019 (UTC)
Well, your rule-of-thumb for reviewing would produce awful results for a 1000 × 8000 letterbox panorama, yet the 8MP rule would permit it. I agree with Cart that worrying about images that fall 10% below the threshold is pointless. There will always be images below any threshold. In 2017 only 5% of images were < 2000px on the shortest side. In 2019, this should be noise. Let's worry more about ensuring QIs really are any good as images. I see MP is gaining most support at present, though I wonder if its supporters have installed the Javascript that reports MP on the file description page. I assume those who support that rule will also roll their sleeve up when it comes to enforcing it and fixing up all the nominations that should not have passed. I think we've just made work for ourselves: MP sounds simpler but in practice it is not. -- Colin (talk) 09:44, 28 January 2019 (UTC)
I will upload fewer widlife images of encyclopaedic value to Wikipaedia projects if the QI theshold is increased. Is that what the community wants? Charles (talk) 11:50, 30 January 2019 (UTC)
Is getting a QI or FP stamp on your photos the only reason for uploading them on this project? It shouldn't be IMO, seeing the photo used in articles should be reward enough. --Cart (talk) 12:29, 30 January 2019 (UTC)
+1 agree. I fail to see why it should be important that all one's uploaded images are tagged with a QI or FP mark. Both should be something to aim for rather than set low. If your 3MP butterfly-in-flight fails to get QI, then you have an incentive to try again to get a bigger photo. Thus either your fieldwork will improve with more practice, or you have an excuse to lust after a bigger lens :-). Or, like the rest of us, just accept that not every useful photo meets the threshold. I would argue that having a higher threshold will encourage people to take and upload better quality and higher resolution images, which, yes, is what the community wants. -- Colin (talk) 13:51, 30 January 2019 (UTC)

┌─────────────────────────────────┘
A related thing about why 'shortest side' might be a good thing, just popped up on Commons talk:Featured picture candidates#FPCBot. --Cart (talk) 16:38, 2 February 2019 (UTC)

Cart you ask: Is getting a QI or FP stamp on your photos the only reason for uploading them on this project? You impugn my motivation, but please examine what I do here. I upload images that are good enough to go in Wikipedia articles. Many will be VI and/or QI and a few FP. Sometimes we already have photos better than mine to illustrate articles, but if I feel my images add to the encyclopaedic content (often being from a different country) I will upload them. I won't bother if the community does not appreciate them. Charles (talk) 11:49, 5 February 2019 (UTC)
And responding to Colin's comments. My fieldwork improves with practice, but since the buttefly lives only in Madagascar, futher practice is impossible. And my fieldwork to date has shown that the only way to photograph this fast-moving swallowtail is using a hard-held lens, so lusting after a bigger lens wouldn't help. Charles (talk) 11:49, 5 February 2019 (UTC)
Charles, you are too hard on yourself. The community appreciates ALL your photos. They are all valuable for articles. Having a QI, VI or FP stamp on them is only a little silver lining. A photo can be very valuable for articles and appreciated even if it has no assessments, simply because it is good enough and one of the very few of a certain subject. Like this one, used on five wikis. And of course your swallowtail. When I write articles, I don't give a rat's ass about assessment when I choose photos to illustrate it, best illustration at thumb always wins. --Cart (talk) 12:10, 5 February 2019 (UTC)
Charles, you claimed you "will upload fewer widlife images of encyclopaedic value to Wikipaedia projects if the QI theshold is increased." You now claim you upload images to illustrate WP articles, and don't bother about community appreciation. So was that first claim simply a bluff, and your "is that what the community wants" just a blackmail threat? Either getting the QI badge influences your decision to upload or not. Perhaps you could indicate which by striking out the comment that is not true. -- Colin (talk) 12:31, 5 February 2019 (UTC)
Colin. As usual you twist what I say. You state that I claim that I don't bother about community appreciation. I never said that. You also accuse me of blackmail. Stop bating me. Charles (talk) 16:06, 5 February 2019 (UTC)
How do you explain what you mean by "I will upload fewer widlife images of encyclopaedic value to Wikipaedia projects if the QI theshold is increased. Is that what the community wants?"? -- Ikan Kekek (talk) 02:50, 6 February 2019 (UTC)
I refer Ikan Kekek to the small number of 'quality' images I currently upload that are not suitable for Wikipedia articles. There may be better pictures there already or the article size may not justify any more. These images may have encyclopaedic value, perhaps because they are from a different country. They may be images of more common animals which have been photographed many times. If they cannot become QI because of size then they will not be found under 'good pictures'. I will not give my images a wider crop just to pass the QI threshold, so I wouldn't bother to upload them. Charles (talk) 16:38, 6 February 2019 (UTC)

Comments part 2[edit]

Charles, I'm sorry to say that almost none of your QIs can be found anyway. If you look at the discussion below regarding #Recently promoted QIs, you will see that the old category system is broken and you have not sorted your photos yourself into the different subsections of Category:Quality images of animals. All your QIs are in the big bag called Category:Quality images. It contains 205,759 images at the moment. I suggest you start adding the right QI categories to your photos if you want people to find and identify them as 'good pictures'. It's a messy situation. --Cart (talk) 19:36, 6 February 2019 (UTC)

I never imagined that is how people search. Is it really Cart? And nobody told me the category system was broken. If I'm looking for lion pics I go to category:panthera leo and start from there. That's where ALL the good images (FP, QI and VI) show up and I'm happy with that. It would be silly to search for QIs your way then have to search again for VIs and yet again for FPs. I will have a look at Category:Quality images of animals though. Never been there. Charles (talk) 19:48, 6 February 2019 (UTC)
It is a shambles. We only need categories, not pages. Charles (talk) 19:59, 6 February 2019 (UTC)
(Edit conflict) Well, it's not "my way", it's just the only way we have at the moment if we want to single out the QIs. The system has been broken for a while now, but nobody has cared about it. That is why the discussion was started a few days ago. Please read it and you might understand. Of course all QIs, Vis and FPs are listed in the normal categories, but it can sometimes be hard to find them there since the selection button to show them is broken so often. We have two systems for categorizing FPs: "Fetured Pictures of..." and the FP gallery. You don't think they are silly since you use both of them. This is the QI equivalent of those. Now we must keep track of the QIs ourselves just like we do with FPs. --Cart (talk) 20:00, 6 February 2019 (UTC)
I use both FP systems but I do actually think it's silly to have two. BTW, I know you said to wait, but I tested the delete cat-a-lot functionality and it doesn't work! I tried to remove Category:Quality imgaes from the two images in Category:Quality images of crabs by Charlesjsharp Charles (talk) 21:29, 6 February 2019 (UTC)
You are right about the QI cat, I tested it on mine and it kicked back. Hmmm, strange, it might be because all QI cats are hidden categories. That is also one of the reason we have two ways of displaying FPs: The gallery is pages that anyone, even unregistered users, can see. A way of displaying our finest images. All the FP (and QI) categories are hidden, that is you have to be logged in to see them plus have the correct preference settings. The QI gallery pages were to showcase the QIs the same way. Not sure we should "unlock" the QI categories and make them visible since they are interconnected with the user categories. A mess. --Cart (talk) 21:46, 6 February 2019 (UTC)
Ok Charles, I found it. You can't remove the parent category Category:Quality images because it also functions as a sort of meta-category linked to the {{QualityImage}} template. If you remove it, the green QI stamp disappears. The subcategories do not fill in for it. So you have to live with having both the parent and the sub cats. --Cart (talk) 22:51, 6 February 2019 (UTC)
Ok thanks. Charles (talk) 10:09, 7 February 2019 (UTC)

"If they cannot become QI because of size then they will not be found under 'good pictures'. I will not give my images a wider crop just to pass the QI threshold, so I wouldn't bother to upload them." This is a misunderstanding of how people find images on Commons. As others note, the system is broken. Categories are hard to use to find things. The "Good images" tool only works some of the time. And since QI is only awarded to user-created content, it is a very very poor way of finding content of quality on Commons: most of Commons images come from elsewhere, like Flickr or museums. If your image is well described, categorised and looks appealing at thumb, then a user will probably have a look at it and may use it. The purpose of QI is not to help users find which images to bother examining -- if it were, we'd open it up to all images. Google will find your picture better than Commons search or category browsing. And btw, a close crop isn't always what users want. It might be useful for a passport photo image on a wiki article, but terrible for a page in a magazine where some caption or title text may be overlaid. -- Colin (talk) 12:26, 7 February 2019 (UTC)

Next steps[edit]

The voting deadline has now passed, so where do we go from here? Option 1 is the only one with majority support. I see three ways we can approach this:

  1. Have someone uninvolved close the discussion. The percentage of support is 60%, just on the cusp of consensus, so I could see this going either way but it'll be up to the closer.
  2. Take a revote with only Option 1 available. Some opposed because they preferred a different option. Faced with the choice of only 2 MP vs. 4 MP, perhaps we can achieve a more conclusive result.
  3. Go back to the drawing board to discuss how to carve out exceptions (e.g. wildlife) to remain on 2 MP, while increasing the threshold for other photos.

Thoughts? -- King of ♠ 04:38, 10 February 2019 (UTC)

  • There were only a few commoners contributing to the vote and even I've missed the voting. Maybe this discussion should have (by whatever means) been made more visible? --Granada (talk) 08:55, 10 February 2019 (UTC)
In the first 10 days of February alone, we had 86 distinct contributors to the QI candidate page. I'd guess then there are well over 100 regular active contributors here. So, to be honest, I think participation on this poll was too low to draw any conclusions about anything other than that perhaps for QI resolution is not a big concern, and that we have no consensus about how/whether to change the threshold. Even for wildlife, we only have arguments from one person, and we can't assume they are representative of any other wildlife photographers. Perhaps someone is able to produce stats like I did for featured pictures. At FP, low resolution pictures have been nearly eliminated naturally, but perhaps this is not true for QI, if the 2MP threshold is still considered "enough" by many photographers or reviewers. I also note my own continued confusion about what QI is for. It seems we get a different answer from each person. Is it here to motivate photographers or here to help image seekers find good content? The latter is compromised when you consider the wider base of pictures on Commons are mostly imported from elsewhere, and so not considered for QI. Perhaps eventually WMF will improve the UI/search for Commons so that a potential user, an image seeker, can easily set their own thresholds for image size when searching or browsing categories.
Wrt polling, it is often warned on wiki to not take a poll unless you feel a discussion has reached consensus. People tend to vote wrt their own personal wants/needs, and it takes a discussion with lots of contributors to consider other viewpoints. People also tend not to want to change their opinions when they've already voted, or express their opinions in more polarising language. I think one problem was this discussion started over the Christmas holidays, when there was less activity, and perhaps like Granada says, it should have been advertised with a banner at QI. Perhaps we could have another discussion later in the year, and if someone has some stats that might indicate the spread of picture sizes being offered/accepted at QI. -- Colin (talk) 11:19, 10 February 2019 (UTC)

Composition criteria at QI[edit]

I seek some guidance on the composition requirements for QI candidates. This has been prompted by voting on this image and this similar image. The oppose vote is from Diego, probably Commons' top photographer, so his oppose has to be taken seriously. The support vote is from Ikan Kekek, not a photogrpaher, but a known objective and thoughtful reviewer. There is nothing in the QI guidelines, as far as I can see, that says these images cannot be QI (assuming they pass focus, lighting etc. requirements). Many successful QI candidates are compositions with bits of car, building, plant or person cut off. Should they be rejected too? Here is one example of an image I just opposed. It had been approved by Ermell. The bottom of the temple steps are missing. Also, is the person in the doorway grounds for rejection? --Charles (talk) 11:34, 16 January 2019 (UTC)

The QI and FP guidelines do not specify how to judge composition. And rightly so because there is no algorithm or threshold. QI is the vote of one person, and some votes are contested. It is quite possible for one photographer to consider cropping a tail to be a flaw and for another to consider it unimportant. Whether it is acceptable that something is cropped or only partially visible is simply a matter of judgement. In this FP nomination, Code asks if a bit more of the bottom could be found as the crop is rather tight. It is rather tight, but reviewers differ as to whether that is enough to prevent their support or so bad to encourage an oppose.
I don't think your response "A bizarre decision yet again from this user" to Poco's "Oppose Bottom crop" is acceptable behaviour at all, Charles. You followed it with your oppose review "Half a car doesn't make a good composition" that contradicts's Poco's support vote on an image. And followed that it a few minutes later with a suspiciously identical "Oppose bottom crop" vote on one of Poco's photographs (that you link above). Are we to consider you were in a neutral frame of mind wrt Poco's images and reviews? Please can you keep personal attacks out of your reviews and responses to reviews. -- Colin (talk)

Btw, the Commons:Photography terms has a number of topics that may help new (and old) reviewers become familiar with terms used at FP and QI. Some of the key reviewing qualities to consider at QI and FP are:

-- Colin (talk) 13:46, 16 January 2019 (UTC)

As usual, you use inflamatory words in reponse to my reasoned request. You say "suspiciously identical"? It was deliberately an identical review which I link to above. It's not suspicious, so don't insinuate it is. It's not revenge either, as I openly opposed using the same criteria to raise a genuine concern. Please can you stop your personal attacks. --Charles (talk) 15:23, 16 January 2019 (UTC)
Charles, I am not using "inflamatory words" nor am I making personal attacks. Once again you seem to be transferring your personal issues onto others. I cannot read your mind, so said "suspiciously similar". Now you confirm they are "deliberately" similar, we can all see your state of mind. So you deliberately chose to use identical words in an opposing review a few minutes after calling Poco's review "A bizarre decision yet again from this user". You are making a personal attack of Poco -- claiming he frequently makes "bizarre decisions". And then offer the same review yourself on his photo? Please read Wikipedia:Do not disrupt Wikipedia to illustrate a point. When you say "I openly opposed using the same criteria to raise a genuine concern" you are in fact being WP:POINTy. In the examples of that Wikipedia guideline, if transferred to Commons, could well be:
  • If someone opposes one of your pictures
  • do explain why the picture meets the quality criteria.
  • do not oppose one of their pictures, giving the same rationale.
Charles, I strongly suggest you take a wikibreak. -- Colin (talk) 16:28, 16 January 2019 (UTC)
  • +1. Charles, you've been at me, then The Photographer, Colin, and now Poco. Who's next? Please take a break, this is not good for the project. Even mild-mannered Famberhorst has tried to calm you down. --Cart (talk) 16:48, 16 January 2019 (UTC)
  • I have withdrawn the review comments and apologise for my review. I did not intend to 'disrupt Wikipedia to illustrate a point'. I was cross. But I ask Colin to refrain from goading me. Please read the tone of my introductory paragraph to Composition criteria at QI and then read the tone of Colin's response. --Charles (talk) 17:12, 16 January 2019 (UTC)
  • I can understand when things get heated because of cultural differences and language barriers, but you are both Scots. Or maybe this is how the old Scottish clan feuds are kept alive. Anyway, pour yourselves a dram and relax now. --Cart (talk) 18:18, 16 January 2019 (UTC)
  • Charles, you are indeed "cross" and thinking bad of others. My first edit was to helpfully fix up your links to the two images. My second edit neutrally answered your query along with more helpful links about composition, etc. It was only when I noticed what you wrote at QI to Poco, and then noticed the two other reviews you made subsequently, that I felt you were not being entirely straight with us. The neutral appearance of your query here hides your unacceptable behaviour at QI last night. That's why my tone changed. And quite rightly you should apologise. Charles, once again you blame others for your mood. I am not "goading" you. This is a mess of your own making. As was the dispute with Cart, The Photographer, etc. Nobody made you write and say what you did. The best way to deal with a cross mood, is to take a break. And, per Cart, I can recommend Bruichladdich's "Port Charlotte" 10-year-old Islay single malt. -- Colin (talk) 18:29, 16 January 2019 (UTC)
Since I was mentioned in this thread: I don't care as much about cutoffs of tails as many other voters here, if I find that the composition is in other ways good. However, I completely respect the fact that others do care about that more. Different conceptions of what constitutes a good composition are, as many Americans might say, what makes ballgames. -- Ikan Kekek (talk) 10:06, 18 January 2019 (UTC)
I have been very busy in real life in the last days, that's why I didn't participate here so far.
Indeed, Colin, I didn't find Charles's reaction really civic when copypasting my review when opposing one of my pictures, right after I opposed one of his. I write this because the topic has been raised, but I don't usually spend my time in letting others know that their edits are not as they should be, that usually ends up in a bad way (blocks, too long breaks, short breaks are sometimes good, though,...) and after long discussions. To me spending time there is just a pity. Anyhow, I appreciate when others find the time and right wording to point such things out, as Colin does.
After seeing those edits I also thought that Charles was being to harsh and probably with a too high wikistress level. I don't think (or at least want to believe) that he has any personal problem with me. We have exchanged some emails some time ago and he has always been very kind, not to mention the high value of this contributions.
No, coming back to the topic of this thread. To me it's indeed disturbing to watch the picture of an animal with a cut off tail, that's not a QI to me. If you concentrated on details that it would be a different thing but an image with 99% of an animal is just wront to me. Poco2 18:56, 18 January 2019 (UTC)
I respect Poco2's opinion on cutting tails off, but point out how hard it is to be harsh on oneself too: 1, 2, 3, 4, 5 -- Charles (talk) 17:31, 6 February 2019 (UTC)
Your comment makes me sad, Charles. In all those 5 examples the whole tail is not visible because it's either hidden behind an object or I focused on the body of the animal. I declined your images because you cropped it unfortunately. Sorry if I don't list a lot of examples, my time is too valuable to me to waste it the way you did with yours --Poco2 21:17, 6 February 2019 (UTC)
No need to be sad. He happy in your belief that this image is QI wheras my chameleon isn't. Charles (talk) 21:37, 6 February 2019 (UTC)

"Self-photographed"[edit]

Please see a discussion about this misleading, confusing piece of bad English at the thread on the QIC nomination of File:2019-01-03 BMW IBSF World Cup Bobsleigh and Skeleton Altenberg StP 0005 LR10 by Stepro-2.jpg, currently the first nomination for January 18, 2019, and at Template talk:Own#Own photo etc., which I copied to Template talk:Self-photographed#Initial discussion on creation of this template. No-one would say in English "that image is self-photographed", and the expression "self-photographed" would be expected to describe a self-portrait and nothing else. I strongly urge the deprecation of this expression. If you want an unambiguous expression, "photographed by me" would do it. Ikan Kekek (talk) 03:48, 21 January 2019 (UTC)

The use of first-person pronouns is frowned upon in formal English; Having "photographed by me" in the "Source" field sounds weird to me. "Self-photographed" sounds fine if it is only used on photos of others' artwork and not photographs of people (it reminds me of the EXIF tag "A directly photographed image"). For everything else there's {{own}}. -- King of ♠ 07:19, 27 January 2019 (UTC)
But in the nomination I cited above, it was a photograph of a bobsledder. And I disagree with you on first-person pronouns being frowned upon in formal English. Passive voice is often preferred in technical reports, though the trend recently, to my knowledge, has been toward using first person ("we"; "the team") in those, too, but first person, rather than "one" or passive voice, is generally preferred in other types of formal English (essays, etc.), and the point is that "photographed by me" (or "photographed by the uploader", if you prefer third person) is absolutely unambiguous. If you'd like to specify in the documentation for the template in question that it should be restricted only to reproductions of existing artworks, that would be a big improvement, though, because it would be used less. -- Ikan Kekek (talk) 07:54, 27 January 2019 (UTC)
What's wrong with "own work" for the bobsledder? The only reason "self-photographed" exists is to avoid ambiguity in the case of a photo of an existing artwork. -- King of ♠ 02:37, 31 January 2019 (UTC)
There is absolutely nothing wrong with "own work", but some people think "self-photographed" superseded that template and is somehow more accurate in every case. -- Ikan Kekek (talk) 08:46, 4 February 2019 (UTC)

Recently promoted[edit]

Daily many images are promoted, and many users are told by the QICbot "We also invite you to take part in the categorization of recently promoted quality images". But nobody do so. So Commons:Quality images/Recently promoted ist overflowing. Is there any interest to continue this categirzation? At least everyone could categorize their own pictures. --Milseburg (talk) 10:35, 2 February 2019 (UTC)

The QI system is outdated and hasn't been functioning well in years. It was ok when it was first set up, as people had no choice but to organize things manually and there weren't so many photos. Today, we are used to having things automated by using scripts and Bots, so organizing manually doesn't come naturally. Also the way the QIs are organized, listed on pages with no file names or caption, instead of in real categories, seems ancient. (Take a look at two "small" pages, Commons:Quality images/Subject/Animated and Commons:Quality images/Subject/Microscopic, for example.) It can't even be called "categorization" any more. The pages listed on COM:QI are so big and full of photos, they are almost impossible to open and no one can find anything on them due to the lack of text. No wonder people are giving up sorting. FPs are organized the same way (except they are at least labeled), but they are not as numerous (11,895 vs 205,579), so a few users can handle the sorting and fixing of new pages to keep the system afloat.
I think it's time to re-structure the whole QI thing. I propose we use the categories instead of pages, that way you get max 200 photos per page and these are indexed and sub-categorized to handle the big bulk of QIs. Categories should be selected during the QIC process, just like with FPs, before the nom is promoted. The Bot task needs to be re-programmed to put the image in the right category.
It will take a bit of work, reorganizing the whole thing, but something has to be done or the QI project will fail. --Cart (talk) 22:44, 2 February 2019 (UTC)
I like this idea of submission-time-categorization (and using real categories), it seems more useful and would make for a more streamline process. I'm already categorizing my QIs using real categories but not doing the suggested page-categories because that would be even more work and it seems to be the least useful option. --Trougnouf (talk) 22:53, 2 February 2019 (UTC)
Something else I thought of before would be to have a bot that automatically looks for the closest parent category that has a "Quality images of" subcategory for each category that's present. I'm not sure how easy that would be to implement and I have no idea how resource hungry that would be. --Trougnouf (talk) 23:00, 2 February 2019 (UTC)

I wasn't really interested in this categorisation system since I´m ative in QIC. But since a few month some categorisation managers gave up their work because of overload. It started bothering me, that under my older images it calls "recently promotet", although there can be no talk of "recently". So I begann categorizing my own ones and some others. It´s not satisfieing. I am dispassionately wether the system will be abolished or changed. I have also the idea, that the potential QI category will be already given at the beginning of the nomination. For technical implementation, I lack the know-how. But the current state should not stay.--Milseburg (talk) 12:01, 3 February 2019 (UTC)

@Milseburg: Unfortunately, Commons:Quality images/Recently promoted is not exactly user friendly, so some users (including me) opt to categorize via another method. The problem is, that QICbot doesn't recognize, if the images have been categorized by another method and keeps images listed. --MB-one (talk) 15:50, 3 February 2019 (UTC)
@MB-one: It´s quite easy and goes quickly to categorize an image. The mass is the problem and that noone do this with his own sucessfull nominatons. What method do you use? So I can stop now categorizing your car-QIs. Maybe they remain "recently promoted" for ever. --Milseburg (talk) 12:50, 4 February 2019 (UTC)
@Milseburg: I assign the relevant QI categories manually together with my QI user category. Since they don't vary much. Thats relatively easy and covers more dimensions (type of object, location, event), than the tool. I'll try to remove those images from the recently promoted page to declutter it. --MB-one (talk) 13:48, 4 February 2019 (UTC)
I stopped working on the page after only a very small number of brave photographers had helped to clear the editing backlog on two previous calls. It is very ignorant that the vast majority of photographers involved in the QIC process obviously have enormous time to do the most elaborate panoramas, stitching, perspective corrections, noise- and CA removal, but not for the daily seven mouse clicks to sort in their successful candidates. It sucks. Translated with www.DeepL.com/Translator --Smial (talk) 16:11, 4 February 2019 (UTC)
Yes, a very sad reality. (same problem at FP) That is why I think it would be better if we could somehow have a system where the nominator had to fill in the correct category, for a Bot to sort, at the same time as nominating, or the image would not be promoted. --Cart (talk) 16:27, 4 February 2019 (UTC)
Categorization before the nomination sounds like a good idea. --MB-one (talk) 21:04, 4 February 2019 (UTC)
+1 --Smial (talk) 22:59, 4 February 2019 (UTC)

I would support abolishing QI galleries altogether. QIs just aren't as special as FPs, no one goes down a QI subpage browsing a bunch of loosely categorized, mostly uninteresting but useful images. In fact, some pages like Commons:Quality images/Subject/Places/Natural structures don't even load anymore (and no one seems to have noticed or cared), yet the bot is still updating them (after those images are manually sorted by humans on the sorting page, what a waste of volunteer labor). QI categories are quite sufficient. -- King of ♠ 01:22, 5 February 2019 (UTC)

No matter what we make out here: Who can modify the nomination modalities or the bot accordingly? --Milseburg (talk) 15:25, 5 February 2019 (UTC)

We could start by asking Dschwen who operates QICbot and see if they can help us out with this. They might also have some ideas about what can be done. We could also ask at Commons:Village pump/Technical. This is a problem that affects a lot of users and one of this site's assessments platforms, so it is rather important that it functions well. --Cart (talk) 15:47, 5 February 2019 (UTC)
  • Agreed with King of Hearts, furthermore galleries are not usefull with a lot of images, this work with a reasonable amount of images such the FP galleries. With a big number of files, the categories are much more usefull. And with little luck, with the Structured Data on Commons, Category:Quality images will even be enough. Christian Ferrer (talk) 17:10, 5 February 2019 (UTC)
QI categories are all that's needed. Pages are a waste of time. Categorize before nomination. Categories for QI should match categories for VI. Charles (talk) 19:57, 6 February 2019 (UTC)
Cart It is easy for me to move all my QIs into new categories from my personal QI categories using cat-a-lot but how does one then batch delete them from the parent category:quality images? Charles (talk) 20:09, 6 February 2019 (UTC)
Charles, you can batch delete from categories using cat-a-lot too by using the "-" in that tool, BUT I would advice you to not remove the parent category yet since we don't know the outcome of all this at the moment. Let it stay for now and we'll see where it will all end up. Removing can wait. --Cart (talk) 20:18, 6 February 2019 (UTC)

QI categories and what to do[edit]

From the above comments it is pretty clear that the QI pages("old categories") are not working and people prefer the normal categorization system. However, just like with FPs, the QI pages are linked to and powering the gallery on the page Commons:Quality images, the page you get to if you click on the QI info on the Main page. That page and gallery is visible to anyone, including not logged in visitors, whereas all QI categories are hidden and you need to be logged in and have the correct preference settings to see them. So abandoning the old pages would cancel out our showcase to the general public about QIs and we might need to find some new way to keep that page running. We also have a huge backlog of QIs that are not in any QI categories, they would require some sorting program. Thoughts? --Cart (talk) 12:18, 7 February 2019 (UTC)

I use Commons mainly as an archive for my work on de.WP. Honestly, I have never surfed properly through our showcase. But I think you have mentioned an important issue and it should be kept alive and functioning. Technically, I'm over-demanded. So we should ask the technicians what they think and what is possible. I ask User:Dschwen. But it seems he is little active at the time. Cart, please contact Commons:Village pump/Technical as you suggested. Does anyone else know?--Milseburg (talk) 10:16, 9 February 2019 (UTC)
Recently I've been working on FP subcategories, and can tell you what a pain it is to get everything right. I can't even imagine what it would involve to fix the QI subcategories. Perhaps we could get rid of all QI subcategories other than Category:Quality images by user (kept mostly as a benefit for us contributors), and rely on FastCCI to automatically compute the intersection between regular categories and QIs. -- King of ♠ 04:30, 10 February 2019 (UTC)
I would be all for such thing IF only the FastCCI worked all the time. In my experience it fails more often than it works. Perhaps we should concentrate on getting that tool to work 100% instead. In that case, how should we present QIs on the Main page, should we skip that too? --Cart (talk) 10:18, 10 February 2019 (UTC)

My suggestion is to drop the QI gallery pages, and not to replace them by QI categories. Really, do not create a parallel category system for QIs! It would be . acrazy maintenance nightmare. Just add a single QI category and enforce proper categorization in the main commons category system. The rely on tools to float QIs to teh top in searches and on category tress (such as the "Good Pictures" button). --Dschwen (talk) 01:16, 11 February 2019 (UTC)

First of all I would suggest, Dschwen stop the bot to categorize the successful QIC into "recently promoted" and we delate "We also invite you to take part in the categorization of recently promoted quality images" from the Template:QICpromoted. --Milseburg (talk) 18:37, 15 February 2019 (UTC)
Just remember to keep the Bot's automatic tagging of files as {{QualityImage}} (and other useful Bot functions) since many users don't know that they should add this. --Cart (talk) 19:43, 15 February 2019 (UTC)

Delisting 'quality' images[edit]

Is there any mechanism for delisting images which have received a Quality Image mark? According to the QI rules (bold: my emphasis):

===== Image page requirements =====

  1. Quality images shall have a meaningful file name, be properly categorized and have an accurate description on the file page in one or more languages. It is preferred, but not mandatory, to include an English description.

Five images have recently ([[Commons:Quality images candidates/Archives January 25 2019|25 January 2019) been promoted:

All five are misidentified, despite being easily identifiable species, and therefore do not meet the specifications of meaningful filenames, proper categorisation, or accurate description. There really needs to be some better quality control here: people working on QI should check identifications are correct before being given QI status. I'll be renaming them all shortly, but that will only affect the files themselves, and not the QI archives, where the incorrect names remain. - MPF (talk) 21:39, 9 February 2019 (UTC)

There are four main QI criteria:
  1. Made by a Commoner;
  2. At least 2 MP;
  3. Image quality/basic photographic merit;
  4. Commons metadata.
In general the first two are the only valid reasons to delist a QI. For 3, it's treated as inherently subjective and we don't rehash such discussions again. For 4 (your case), it is an easily fixable problem so for me, if someone has the time to nominate something for delisting, they have the time to fix the problem themselves. -- King of ♠ 21:50, 9 February 2019 (UTC)
It's just taken me half an hour to rename the above 5 images plus a couple of others from the day after: it is a slow process when the File Renaming has to do several pages that the image is used on, and then each file has to be edited 'manually' to correct the description. That's time taken away from other Commons editing, and no, I don't have additional time to fix the problem otherwise. - MPF (talk) 22:07, 9 February 2019 (UTC)
Is it the end of the world if an image with bad categories has a QI badge on it until someone comes by to fix it? As w:Template:Afd-merge to says, "If you find that such action has not been taken promptly, please consider assisting in the merger instead of re-nominating the article for deletion." -- King of ♠ 23:06, 9 February 2019 (UTC)
I just don't understand the need to go to QI pages and place the new file name on them. QIs are re-named and re-categorized all the time, that is why we have automatically created redirects. If a file has a faulty name, you move it and change the categories, that's it. It's only if the file is used in Wikipedia articles or if they are in the FP gallery (which is more specific than the QI pages) that you have to correct things. --Cart (talk) 23:48, 9 February 2019 (UTC)
OK thanks - I must have been thinking of the FP situation (re "which is more specific than the QI pages") - MPF (talk) 02:06, 10 February 2019 (UTC)