Commons:Village pump/Technical
Village pump/Technical |
Bug reports |
Code review |
Tools |
Tools/Directory |
Idea Lab |
This page is used for technical questions relating to the tools, gadgets, or other technical issues about Commons; it is distinguished from the main Village pump, which handles community-wide discussion of all kinds. The page may also be used to advertise significant discussions taking place elsewhere, such as on the talk page of a Commons policy. Recent sections with no replies for 30 days and sections tagged with {{Section resolved|1=--~~~~}} may be archived; for old discussions, see the archives; recent archives: /Archive/2024/04 /Archive/2024/05.
- Feature or bug reports should be filed on Phabricator (see how to report a bug). Bugs with security implications should be reported differently (see how to report security bugs).
- Have you read the FAQ?
SpBot archives all sections tagged with {{Section resolved|1=~~~~}} after 1 day and sections whose most recent comment is older than 30 days. | |
Cat-a-lot does not work for categories[edit]
There seems to be a technical problem with Help:Gadget-Cat-a-lot: when I click on Preferences, click on "Allow categorising pages (including categories) that are not files", then select one or more subcategories and try to move/copy/delete them to/from another category (like I always do), nothing happens but an endless "Editing page 1 of X". Moving files is no problem. Can this be fixed? JopkeB (talk) 11:05, 16 February 2024 (UTC)
- I am not the only one having troubles, see Commons:Village pump#Cat-a-lot as well. JopkeB (talk) 13:27, 16 February 2024 (UTC)
- Same for me: moving categories does not work, moving files works. --Reinhard Müller (talk) 14:49, 16 February 2024 (UTC)
- I confirm that it is not working. Theklan (talk) 09:36, 24 February 2024 (UTC)
- I can also confirm that it's not working for categories, although it will sometimes say that the category was moved.Smasongarrison (talk) 00:28, 27 February 2024 (UTC)
- I confirm that it is not working. Theklan (talk) 09:36, 24 February 2024 (UTC)
- Same for me: moving categories does not work, moving files works. --Reinhard Müller (talk) 14:49, 16 February 2024 (UTC)
- Note: If there is specific categories where problem can be replicated then please list them. (ie. fixing requires example cases) --Zache (talk) 15:58, 16 February 2024 (UTC)
- I would like to copy a lot of subcategories in Category:Energy by type of energy to Category:Energy by topic (all except for the 'by' categories). JopkeB (talk) 16:20, 16 February 2024 (UTC)
- It was broken by a change to the HTML structure of the category pages, see phab:T355636#9553616. I hope it will be fixed by another change to the HTML structure, but if not, an interface admin should update the method getMarkedLabels() to account for the change. —Tacsipacsi (talk) 14:02, 18 February 2024 (UTC)
- Same problem for me. I have posted this on MediaWiki talk:Gadget-Cat-a-lot.js but the only replies are people saying "Same problem for me". So nobody is actually looking after this? Cnbrb (talk) 23:19, 27 February 2024 (UTC)
- I would like to copy a lot of subcategories in Category:Energy by type of energy to Category:Energy by topic (all except for the 'by' categories). JopkeB (talk) 16:20, 16 February 2024 (UTC)
For me moving categories was never possible with cat a lot. Always moved one by one.Paradise Chronicle (talk) 05:24, 14 March 2024 (UTC)
Question Is there any chance that we get an indication of:
- What is the priority to solve this problem?
- Whether someone is working on the solution of this problem?
- How long it might take before it is solved?
--JopkeB (talk) 07:13, 5 April 2024 (UTC)
- the solution would be to revert phab:T355636. Can someone do that? Enhancing999 (talk) 15:25, 17 April 2024 (UTC)
- No we can just apply this change to Cat-a-lot, as Tacsipacsi indicated above. Nardog (talk) 16:03, 23 April 2024 (UTC)
- Can someone actually implement the fix? Smasongarrison (talk) 22:03, 23 April 2024 (UTC)
- Only an interface admin can. Nardog (talk) 20:40, 24 April 2024 (UTC)
- Nardog has asks on https://commons.wikimedia.org/wiki/MediaWiki_talk:Gadget-Cat-a-lot.js#Cat-a-lot_failing_202402 to solve this problem. Thanks! One big step forward. JopkeB (talk) 04:08, 25 April 2024 (UTC)
- Only an interface admin can. Nardog (talk) 20:40, 24 April 2024 (UTC)
- Can someone actually implement the fix? Smasongarrison (talk) 22:03, 23 April 2024 (UTC)
- No we can just apply this change to Cat-a-lot, as Tacsipacsi indicated above. Nardog (talk) 16:03, 23 April 2024 (UTC)
Mobile versions of user talk pages are useless[edit]
Though I'm not active on Commons, I got bored and used my mobile device to look through some deletion requests, then at at a user's talk page. I knew that user had several deletion requests, but saw only one on their talk page. The talk page's history showed that old sections were archived, which was fine, but the page seemed to have no links to those archives.
I couldn't find the 'Read as wiki page' (or whatever it is) option, so out of desperation, I switched to the desktop version of the page. I was shocked at what I found there: not only the archive links, but also a notice that the user had been blocked indefinitely. (Before that, I hadn't seen any reference to a block.)
In summary, the mobile versions of user talk pages (and perhaps talk pages in general) are so crippled as to be useless. Brianjd (talk) 08:22, 4 April 2024 (UTC)
- Yes, the mobile experience has always been really bad and as someone who basically exclusively edits with a mobile device it's not a fun experience. The worst part is that in most of the developing world the vast majority of people who surf the web do so using a mobile device. Not too long ago the experience of talk pages on mobile devices wasn't this bad, but they basically made it worse during an update a few months ago. After many, many years of neglecting the mobile experience I sincerely doubt that the developers are planning on fixing it anytime soon. --Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 10:47, 9 April 2024 (UTC)
- should mobile view on Commons be deactivated? Enhancing999 (talk) 15:18, 17 April 2024 (UTC)
- That might go to far, to be honest, the best solution would be to make the mobile version identical to the desktop version but tailored for smaller screens, though the Wikimedia Foundation (WMF) have confirmed multiple times that they will not do that. I have no idea why they insist on delivering an inferior and often infuriating experience to us mobile users, but there doesn't seem to be much we can do about it. De-activating it would break more than it would fix. --Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 15:44, 17 April 2024 (UTC)
- I wonder what there is to break. Consider file description pages: mobile version lacks have of the content (categories).
- I think the skin running on each WMF site is determined by the communities. Accordingly, e.g. dewiki doesn't use the same as enwiki, as they consider it not suitable. Enhancing999 (talk) 15:49, 17 April 2024 (UTC)
- @Enhancing999 If you login in mobile view, select settings from top right menu and then toggle
Advanced mode
to be enabled then the categories are visible. --Zache (talk) 17:37, 17 April 2024 (UTC)- I don't think it's available without login. We could keep the default mobile view from logged in users who require it (e.g. WMF staff for testing purposes). Enhancing999 (talk) 18:14, 17 April 2024 (UTC)
- Yes, this "solution" only works if you have a Wikimedia SUL account and happen to know where to find the button to enable it. Imagine you're a re-user looking for similar images (or other files) from a certain category and can't figure out how to do that. Most people surf the web on mobile devices these days, this is especially true for people from developing countries. When I was in the rural Philippines a few weeks ago I met several people who have never worked with or used a laptop or desktop computer in their lives but all use smartphones, even homeless people have smartphones, these are the most affordable internet-connected devices for most people in the world. Yet, for whatever reason "the mobile experience" of Wikimedia websites is beyond bad. If possible we should put it to a vote to show the MediaWiki developers that we don't want to suppress vital information for a major part of the audience. Files are for re-users and if they can't find the files then there's something seriously wrong with the system. --Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 18:34, 17 April 2024 (UTC)
- I have been impression that "advanced mode" is for things in beta testing, not ready for primetime, for whatever reason. We could though ask what are plans to primetime it. --Zache (talk) 18:40, 17 April 2024 (UTC)
- In mobile view parent categories aren't even visible. How useful can that be? Enhancing999 (talk) 19:34, 17 April 2024 (UTC)
- @Donald Trung, Please, provide the link here to any question "put to a vote." Thanks, -- Ooligan (talk) 17:49, 23 April 2024 (UTC)
- I have been impression that "advanced mode" is for things in beta testing, not ready for primetime, for whatever reason. We could though ask what are plans to primetime it. --Zache (talk) 18:40, 17 April 2024 (UTC)
- Yes, this "solution" only works if you have a Wikimedia SUL account and happen to know where to find the button to enable it. Imagine you're a re-user looking for similar images (or other files) from a certain category and can't figure out how to do that. Most people surf the web on mobile devices these days, this is especially true for people from developing countries. When I was in the rural Philippines a few weeks ago I met several people who have never worked with or used a laptop or desktop computer in their lives but all use smartphones, even homeless people have smartphones, these are the most affordable internet-connected devices for most people in the world. Yet, for whatever reason "the mobile experience" of Wikimedia websites is beyond bad. If possible we should put it to a vote to show the MediaWiki developers that we don't want to suppress vital information for a major part of the audience. Files are for re-users and if they can't find the files then there's something seriously wrong with the system. --Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 18:34, 17 April 2024 (UTC)
- I don't think it's available without login. We could keep the default mobile view from logged in users who require it (e.g. WMF staff for testing purposes). Enhancing999 (talk) 18:14, 17 April 2024 (UTC)
- @Enhancing999 If you login in mobile view, select settings from top right menu and then toggle
- That might go to far, to be honest, the best solution would be to make the mobile version identical to the desktop version but tailored for smaller screens, though the Wikimedia Foundation (WMF) have confirmed multiple times that they will not do that. I have no idea why they insist on delivering an inferior and often infuriating experience to us mobile users, but there doesn't seem to be much we can do about it. De-activating it would break more than it would fix. --Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 15:44, 17 April 2024 (UTC)
Is there any way to get a list of users who uploaded images which appear in a Commons category?[edit]
Hi all
I'm working on some documentation around organising photography events and competitions and would really like to find a way to do this e.g I want to get a list of all the user names of uploaders for all the images in the category and all subcategories for Category:Potatoes. It would be a way to understand who is interested in the topic. I have a feeling this info might also be really useful for people who run the Wiki Loves photo competitions as well e.g 'this many people took part this year' although they may have other ways to do this.
Any suggestions would be really appreciated.
Thanks very much
John Cummings (talk) 09:16, 8 April 2024 (UTC)
- @John Cummings: It’s possible using the API: query+imageinfo can be used to get uploaders from a given category (but not subcategories), and query+categorymembers can be used to get the direct subcategories of a category. I don’t know about any ready-to-use user-friendly solutions. —Tacsipacsi (talk) 19:33, 14 April 2024 (UTC)
New tool for detecting logos[edit]
Hi all! As you already probably know, the Structured Content team is working this year on improving the current user experience with UploadWizard. We already have done some work on the “release rights” step, and we recently concluded a community discussion about the “describe” step. We are currently integrating the feedback received from you into our workflow.
Another thing we are working on is a potential improvement to automatically detect logos when uploaded on Commons through UploadWizard, in order to facilitate their evaluation by the community. A need for machine detection tools was raised in several discussions and user interviews we had in the past with the community, and logos are the second reason for media deletion after Freedom of Panorama, so we decided to work in that direction.
The tool we developed has shown promising results (accuracy is ~96%); in case you’re interested, you can see a brief summary of an evaluation of a sample batch of images. Our intention is for you to discuss and then, if consensus is reached, to integrate the tool in UploadWizard, in a way that would be beneficial for moderation workflow.
We would love to have your input. Do you think this tool could be useful? Do you think this tool could be integrated in UploadWizard, and then integrated in your moderation workflow? Sannita (WMF) (talk) 10:18, 9 April 2024 (UTC)
- Maybe you could run it through the "icons" category to test. Enhancing999 (talk) 10:53, 9 April 2024 (UTC)
- If integrated into the MediaWiki Upload Wizard, how would it work? Would it prevent the uploading of a PD-textlogo or would it detect if an incompatible license or attribution is present? Or would it simply add them to a daily page for community evaluation, akin to "User:Minorax/PD textlogo/2020 June 6"? --Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 11:03, 9 April 2024 (UTC)
- @Donald Trung This is actually something that we wanted to discuss with the community. The tool, as of now, only recognises logos with a very high accuracy, but we want to ask the community what to do next. We are open to suggestions. Sannita (WMF) (talk) 11:17, 9 April 2024 (UTC)
- Sannita, I think that it might be wise to index all logos by date of upload on a single page and have the tool also detect the licenses, and then add these into sections like "Logos with Creative Commons licenses tagged as Own work", "Logos with public domain licenses", "Logos with Creative Commons licenses attributed to an external source". That way we can easily go through each type of upload, a lot of (new) users upload free logos as "Own work" with wrong licenses, these are the most problematic, but those uploaded with external links and / or specific licenses tend to be less problematic, so we would immediately know which areas to focus on, but still evaluate the other logos. For example, a very complex logo shouldn't be in "PD-textlogo" and can be deleted if not found to have a free license, which would also be easily detected using such a system. -- — Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 11:26, 9 April 2024 (UTC)
- For clarification "I think that it might be wise to index all logos by date of upload on a single page and have the tool also detect the licenses" would be a page like "Commons:Image detection/Logos/2024/November/12", I deliberately made logos as a sub-category of image detection because of the suggestions below by user "Adamant1" to also add this for postcards. I think that a tool like this could supersede the groups based on category done by the OgreBot today in the future. The page "Commons:Image detection" could be a central hub where reviewers (in the broadest sense) can use the AI-powered tool to find and detect logos. Heck, maybe in the distant future it can also detect images based on freedom of panorama (exterior images) and many other categories. Logos is a good start, though we should make sure that we don't discourage people from uploading, we should simply make it easier to detect possible copyright ©️ violations. We also used to have a page for uploads by new users, but I am not sure if we still have that (I think that I read somewhere that an update to the MediaWiki software made it difficult to maintain), perhaps the "Image detection" page can also have a sub-category for new users as well.
Heck, maybe we can even use this tool retroactively to group images together on a page like "Commons:Image detection/Logos/2012/December/8" and have like a button that trusted users can press to mark an image on that page as "patrolled" independent from the current solution we have. There are many ways that such a tool could be implemented, however, I sincerely hope that it won't be included before an image is published, as we could be missing out of valuable uploads because a new user (or a Wikipedian not familiar with this website) would be scared off by a warning message. -- — Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 13:21, 9 April 2024 (UTC)
- For clarification "I think that it might be wise to index all logos by date of upload on a single page and have the tool also detect the licenses" would be a page like "Commons:Image detection/Logos/2024/November/12", I deliberately made logos as a sub-category of image detection because of the suggestions below by user "Adamant1" to also add this for postcards. I think that a tool like this could supersede the groups based on category done by the OgreBot today in the future. The page "Commons:Image detection" could be a central hub where reviewers (in the broadest sense) can use the AI-powered tool to find and detect logos. Heck, maybe in the distant future it can also detect images based on freedom of panorama (exterior images) and many other categories. Logos is a good start, though we should make sure that we don't discourage people from uploading, we should simply make it easier to detect possible copyright ©️ violations. We also used to have a page for uploads by new users, but I am not sure if we still have that (I think that I read somewhere that an update to the MediaWiki software made it difficult to maintain), perhaps the "Image detection" page can also have a sub-category for new users as well.
- My own humble suggestion is to disable and dismantle this whole thing, and rather redirect WMF’s funding and priorities to actually useful things requested by the community — like, say, fixing the 10 year old bugs that still plague the unified login process.
- I have nothing against some random AI outfit making use of Commons’s publicly available data to “train” their GIGO machines: Good luck with that. But WMF funding should not be used for this sort of useless, misleading, hyped-up nonsensical makework.
- And I know that this suggestion will never be given any attention, but that’s how it goes. -- Tuválkin ✉ ✇ 01:14, 10 April 2024 (UTC)
- @Tuvalkin, I'm not sure I'd have ranked it as the top priority (just since the old wizard wasn't especially bad), but I do think that improving the wizard is a reasonable place to focus. We're a media repository, which makes the process for uploading media a pretty essential function. We lose an untold number of contributions (Sannita, do you have any data?) from people who aren't able to go through it successfully, and it's the first line of defense to prevent/fix problematic contributions (which reduces moderation work later). Sdkb talk 04:48, 10 April 2024 (UTC)
- Oh, I don’t think the upload wizzard should exist, at all. So there’s that. (Before you ask, I use Special:Upload or external tools, like Vicuña and Commonist.) I agree that uploading media a pretty essential function — therefore gamified uploads from clueless jokers should be discouraged. You however both bemoan those lost uploaders (can it be lost, that which never existed?…) and also call for a better ratmaze to make sure their “work” is not too deterimental to Commons, in terms of scope and copyright blunders.
- At the same time many of the same people who cheer for this sort of continued addition of even more bells and whistles to the upload process, supposedly to avoid that loss of untold poor dears, are the same who constantly decry mass uploads, and who are happy to dance on the grave of the most prolific uploader (both quantity and quality) Commons ever had.
- So, basicly not impressed. Wont help the Upload Wizzard this adding to the loop of an opaque step which likely takes pre-published media files away from WMF purview off to some blackbox; the environmental impact of the additional computing resources needed — that’s a cherry on top. (Remings me of: Hop on a jet plane to join all the WMF cronies for needless face-to-face meetings: The cafeteria is 100% vegan because environment, dontcha know?)
- To clarify: In my opinion, one-off uploads, especially by editors creating articles in other projects, should be allowed into Commons via a pipeline that’s not the same of mass or “expert” uploading — I presume the Visual Editor has such a function (never used it — unsurprisingly, I think it’s also nonsense). However, it’s not helping this the training of an A.I. (and I’m all out of irony quote marks at this point) who will nag Clippy-style said unexperienced uploader, muddling the process even more. Especially since it will, most of the time, be confusing for logos all kinds of icons, diagrams, maps, flags, and traffic signs.
- Was I the only one who shook their head and palmed their face at the apparent fact that the control group for this A.I. training was the contents of Category:Logos, with depth=0…? Yes, that one garbage-bag category where end up stuck all the poorly categorized logos, including, more than any deeper subcat, miscurated images which are anything but logos! Just wow — the more I think of this, the worse it looks like.
- But you’re right that this sort of nonsense is (also) «requested by the community». Sad…!
- -- Tuválkin ✉ ✇ 06:39, 10 April 2024 (UTC)
- Re training using a depth of 0, oof — Sannita, you all should rethink that.
- Re your braoder point, you seem to be pursuing a world in which it's far harder for newcomers to contribute to Wikimedia. That's a world in which our projects have more systemic bias and less overall content, which is not what I'd want. Sdkb talk 14:17, 10 April 2024 (UTC)
- @Tuválkin: thanks for your comments. Logo samples in the evaluation dataset were collected via a PetScan query with a category depth > 0. I'm sorry I couldn't retrieve the exact depth from that query.
- Cheers, MFossati (WMF) (talk) 16:55, 11 April 2024 (UTC)
- I agree to Tuválkin’s point that this looks like very weird priorities. Recently, UploadWizard became so broken for me that I have to look into different tools for upload. (I don’t know if it would still be usable with high-income country bandwidth and no power cuts.) Why do people start to work on new features of something whose basic function is more or less broken? Just because AI sounds fancy or it’s interesting to play around with it?
- Just to be clear: Also agree with Sdkb that I’m not in favor of a world in which it's far harder for newcomers [and] more systemic bias. However, this doesn’t invalidate wondering about reasons for prioritizing a potentially useful feature over basic functionality. —Marsupium (talk) 01:25, 7 May 2024 (UTC)
- @Tuvalkin, I'm not sure I'd have ranked it as the top priority (just since the old wizard wasn't especially bad), but I do think that improving the wizard is a reasonable place to focus. We're a media repository, which makes the process for uploading media a pretty essential function. We lose an untold number of contributions (Sannita, do you have any data?) from people who aren't able to go through it successfully, and it's the first line of defense to prevent/fix problematic contributions (which reduces moderation work later). Sdkb talk 04:48, 10 April 2024 (UTC)
- Sannita, I think that it might be wise to index all logos by date of upload on a single page and have the tool also detect the licenses, and then add these into sections like "Logos with Creative Commons licenses tagged as Own work", "Logos with public domain licenses", "Logos with Creative Commons licenses attributed to an external source". That way we can easily go through each type of upload, a lot of (new) users upload free logos as "Own work" with wrong licenses, these are the most problematic, but those uploaded with external links and / or specific licenses tend to be less problematic, so we would immediately know which areas to focus on, but still evaluate the other logos. For example, a very complex logo shouldn't be in "PD-textlogo" and can be deleted if not found to have a free license, which would also be easily detected using such a system. -- — Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 11:26, 9 April 2024 (UTC)
- @Donald Trung This is actually something that we wanted to discuss with the community. The tool, as of now, only recognises logos with a very high accuracy, but we want to ask the community what to do next. We are open to suggestions. Sannita (WMF) (talk) 11:17, 9 April 2024 (UTC)
- This looks like a super cool tool. I was just wishing there was something similar for detecting images of postcards. I have to agree with Donald Trung that it's probably better to use an indexer of exiting files instead of being directly integrated into UploadWizard. I imagine things like this are going to be the future of detecting and organizing specific types of images anyway and I doubt every use case going forward could be integrated into UploadWizard, but a hub for different types of detected images going forward would be great, starting out with logos and then integrating it with other types in the future. I don't think turning UploadWizard into a metaphorical Swiss Army knife of image detection at the point of upload would really be practical or useful though. From what I've seen most people are usually turned off by that sort of thing. Especially new users. UploadWizard is hard enough to understand and work with as it is already. --Adamant1 (talk) 12:57, 9 April 2024 (UTC)
- This seems like a really useful tool! As far as integration into the Upload Wizard, we could use it to customize the user experience. For instance, there could be a dialogue "This looks like a logo. Is that correct? [Yes] [No]". [Yes] would autocategorize the image and lead to follow-up steps trying to confirm the copyright (e.g. directing them to COM:RELGEN if they claim it's their own work), and [No] would apply a hidden category or add it to a feed of reported false positives. Sdkb talk 16:29, 9 April 2024 (UTC)
- Very nice! The model seems to detect reliable for graphics vs. random photos. Let me suggest some curveballs, from the type of images I often encounter: How does it perform when you include Flags and CoAs into the mixup? Uploads of those categories are also often deleted, but less often than logos. Another test could include Map details (i.e. cutouts) which are usually not deleted.
- Also, what would this tool actually do after the detection? Bring the positives into a dedicated (hidden) category for users to evaluate? Yes, I could really see a use in that, for quicker processing of uncategorized uploads. --Enyavar (talk) 12:19, 9 April 2024 (UTC)
- @Enyavar Thanks for your opinion. For now, we limited ourselves to logos because they were easier to detect, but if there is consensus for it, we can expand our work to other kinds of images.
- About what the tool can actually do... that's the point of the discussion. We have suggestions to make, but we want to hear from you first, not to influence the discussion. The dedicated category/tag was actually one of the suggestions that we had, to be fair. But we'll take note of this suggestion, thanks! Sannita (WMF) (talk) 14:12, 9 April 2024 (UTC)
- Ah, sorry, I meant something else with my first comment: can the tool reliably detect logos in the presence of these other (similar looking) images? What are the percentages for examples A, B, C, D, E or possibly F, G, H? It won't be totally bad if these get detected as logos with a certainty of 90%+... I'm just curious whether or not files from these categories have a higher false positive rate for the current tool. --Enyavar (talk) 15:08, 9 April 2024 (UTC)
- Good idea, to test on coats of arms, flags and maps in addition to icons I mentioned above.
- Wonder if the training set was useful: Commons mainly has simple logos for which it doesn't actually matter that they are also logos. Enhancing999 (talk) 20:04, 9 April 2024 (UTC)
- Hi @Enyavar and Enhancing999: {{u}} here, I'm the main technical point of contact behind this effort. Thanks for your feedback, very useful and appreciated! Here are the probability scores of the images you mentioned:
- A few observations:
- all scores but D and E seem low enough to be cut out;
- as a human, I'd consider D and E as logos if I didn't know what coat of arms are. The model was indeed trained on logos and non-logos, so it hasn't got any notion of coat of arms;
- probability threshold selection will be crucial to tune the amount of what we’d like to consider as true positives.
- Anyway, I definitely think that your suggestions make a lot of sense: testing the robustness of the model against visually similar inputs is now in my to-do list.
- Cheers, MFossati (WMF) (talk) 11:03, 10 April 2024 (UTC)
- Ah, sorry, I meant something else with my first comment: can the tool reliably detect logos in the presence of these other (similar looking) images? What are the percentages for examples A, B, C, D, E or possibly F, G, H? It won't be totally bad if these get detected as logos with a certainty of 90%+... I'm just curious whether or not files from these categories have a higher false positive rate for the current tool. --Enyavar (talk) 15:08, 9 April 2024 (UTC)
- This doesn't seem un-useful, but I am a bit skeptical that it will move the needle much in terms of checking logos for copyvio/spam. There is already an overwhelming backlog of this stuff -- 118,798 files in Category:Unidentified logos alone. The most likely outcome is another backlog. Gnomingstuff (talk) 20:42, 9 April 2024 (UTC)
- What would probably make it a lot more useful is if it was used for flow control in the upload wizard. If an image is detected as a logo and the user confirms that it's a logo, that could eventually shunt the user into a workflow that asks them for logo-specific information, e.g. "what does this logo stand for", "where did you find this logo", etc. Omphalographer (talk) 18:57, 10 April 2024 (UTC)
- I can see both sides and arguments here. Tuvalkin is strictly against adding such "another whistle and bell" to the upload wizard (I find it tedious too, and I can understand how a Clippy-style intervention during the upload would frustrate the more experienced uploaders). On the other hand, this addition to the wizard could immensely enrich the descriptions of uploaded logos by one-time contributors (often by the companies/brands who hold copyright of the logo, i.e. these are officially sanctioned uploads that we'll be able to use) who we can never contact again even a week later for clarification on their uploads. --Enyavar (talk) 09:32, 11 April 2024 (UTC)
- What would probably make it a lot more useful is if it was used for flow control in the upload wizard. If an image is detected as a logo and the user confirms that it's a logo, that could eventually shunt the user into a workflow that asks them for logo-specific information, e.g. "what does this logo stand for", "where did you find this logo", etc. Omphalographer (talk) 18:57, 10 April 2024 (UTC)
Thanks all for your interventions! We've already identified some good things we can work on, even though we would like the discussion to go on and let you continue have your say. We would like the discussion to go on some more days, to try to get some consensus going on. Sannita (WMF) (talk) 10:17, 15 April 2024 (UTC)
New tool for cropping and rotating images (proposal)[edit]
MediaWiki or at least Commons should have a function that allows to crop images and save them as separate ones directly, including the original description.
Why would we need it when most phones can crop or even rotate images?
- preserve image resolution
- preserve metadata
- ensure derivation is annotated on original file
We currently have a bot and a user tool that have some of the functionality, but this lags behind a reasonable functionality. Maybe WMF development can be facilitated by adopting some of their code. Enhancing999 (talk) 08:13, 12 April 2024 (UTC)
- Enhancing999, CropTool has no lag for me. You ask to crop an image, you crop it and it is immediately uploaded. I use it regularly and I do not have any complains about it. Rotatebot also seems to work fine. You tag a file for rotating and within 24 hours it is rotated. I guess we could have a tool to rotate immediately, but the current approach is not bad. --Jarekt (talk) 00:42, 18 April 2024 (UTC)
- There are two years of open bug reports and feature requests on the tools talk page.
- I agree it works on files of 500 KB, but larger ones fail now, especially when attempt to fix rotation by a few degrees. Rotatebot can't handle that. Also the automated border detection fails. Enhancing999 (talk) 08:15, 18 April 2024 (UTC)
- Lots of miscellaneous failures over the last few years, and no one committed to maintaining it.
- I'd be glad to see a group of three or more volunteers take it on, or WMF, but finding one more individual to do it just postpones collapse, because the environment keeps changing. - Jmabel ! talk 12:46, 18 April 2024 (UTC)
- Isn't development of key features meant to be done by paid WMF staff? It's not a content/editorial issue. Enhancing999 (talk) 07:41, 20 April 2024 (UTC)
- @Enhancing999:
- Why would we need it when most phones can crop or even rotate images?
- That assumes that users use phones. The section ‘Mobile versions of user talk pages are useless’ did point out that some users exclusively use phones, shunning other devices such as laptops and desktop computers.
- But not all users are like that. There are good reasons for using other devices, including (but not limited to) the criticisms of Wikimedia's mobile versions given in that same section. Brianjd (talk) 08:30, 23 April 2024 (UTC)
- This section is actually about two distinct actions:
- Edit an image (by cropping or rotating), while otherwise preserving image quality and metadata.
- Upload the edited image with an appropriate filename and description and ensure that the original image’s description is updated accordingly.
- But this section proposes to bundle these actions into one tool.
- The problem is that there are other important types of edits besides crops and rotations. Even if Commons can’t help with the edits themselves, it should still help with filenames and descriptions, the same way as it does for crops and rotations. Brianjd (talk) 08:46, 23 April 2024 (UTC)
- Well, in the past, bundling these actions have been proven fairly helpful. If you think other tools are needed at Commons, please discuss this elsewhere. Enhancing999 (talk) 14:35, 23 April 2024 (UTC)
- @Enhancing999: I agree that bundling these actions into one tool is helpful. But that one tool should be implemented as a wrapper around more specific tools (one for cropping and one for handling filenames and descriptions), which should also be accessible separately. Brianjd (talk) 08:21, 24 April 2024 (UTC)
- As user, I'm mainly interested in a stable, working tool. For the rest, WMF pays IT staff. Enhancing999 (talk) 08:40, 24 April 2024 (UTC)
- @Enhancing999: I agree that bundling these actions into one tool is helpful. But that one tool should be implemented as a wrapper around more specific tools (one for cropping and one for handling filenames and descriptions), which should also be accessible separately. Brianjd (talk) 08:21, 24 April 2024 (UTC)
- Well, in the past, bundling these actions have been proven fairly helpful. If you think other tools are needed at Commons, please discuss this elsewhere. Enhancing999 (talk) 14:35, 23 April 2024 (UTC)
- There has been some code written to include cropping into mediawiki. Just needs someone to take it over the line. Bawolff (talk) 21:48, 11 May 2024 (UTC)
Tech News: 2024-16[edit]
Latest tech news from the Wikimedia technical community. Please tell other users about these changes. Not all changes will affect you. Translations are available.
Problems
- Between 2 April and 8 April, on wikis using Flagged Revisions, the "Reverted" tag was not applied to undone edits. In addition, page moves, protections and imports were not autoreviewed. This problem is now fixed. [1][2]
Changes later this week
- The new version of MediaWiki will be on test wikis and MediaWiki.org from 16 April. It will be on non-Wikipedia wikis and some Wikipedias from 17 April. It will be on all wikis from 18 April (calendar). [3][4]
- Default category sort keys will now affect categories added by templates placed in footnotes. Previously footnotes used the page title as the default sort key even if a different default sort key was specified (category-specific sort keys already worked). [5]
- A new variable
page_last_edit_age
will be added to abuse filters. It tells how many seconds ago the last edit to a page was made. [6]
Future changes
- Volunteer developers are kindly asked to update the code of their tools and features to handle temporary accounts. Learn more.
- Four database fields will be removed from database replicas (including Quarry). This affects only the
abuse_filter
andabuse_filter_history
tables. Some queries might need to be updated. [7]
Tech news prepared by Tech News writers and posted by bot • Contribute • Translate • Get help • Give feedback • Subscribe or unsubscribe.
MediaWiki message delivery 23:26, 15 April 2024 (UTC)
changes to Module:Autotranslate[edit]
Module:Autotranslate was altered today to use more efficient test for presence of a language subtemplate. The new version was tested extensively, but please be on a lookout for any issues with templates using it. Please ping me if any issues are found or suspected. Jarekt (talk) 13:34, 18 April 2024 (UTC)
Deletions on "structured tab"[edit]
Anybody know if the feature is still being maintained? I reported a problem a while back about the impossibly to remove wrong statements, but it's still broken. Enhancing999 (talk) 07:44, 20 April 2024 (UTC)
- Just to confirming. Your are referring to the coordinate deletion problem described in phab:T313638 or is there some other bug? -- Zache (talk) 16:08, 20 April 2024 (UTC)
- It's the problem noted here in January 2023 (more than a year ago): Commons:Village_pump/Technical/Archive/2023/01#Delete_wrong_coordinates_not_working. Any idea why nothing happened? Enhancing999 (talk) 14:40, 23 April 2024 (UTC)
Tech News: 2024-17[edit]
Latest tech news from the Wikimedia technical community. Please tell other users about these changes. Not all changes will affect you. Translations are available.
Recent changes
- Starting this week, newcomers editing Wikipedia will be encouraged to try structured tasks. Structured tasks have been shown to improve newcomer activation and retention. [8]
- You can nominate your favorite tools for the fifth edition of the Coolest Tool Award. Nominations will be open until May 10.
Changes later this week
- The new version of MediaWiki will be on test wikis and MediaWiki.org from 23 April. It will be on non-Wikipedia wikis and some Wikipedias from 24 April. It will be on all wikis from 25 April (calendar). [9][10]
Future changes
- This is the last warning that by the end of May 2024 the Vector 2022 skin will no longer share site and user scripts/styles with old Vector. For user-scripts that you want to keep using on Vector 2022, copy the contents of Special:MyPage/vector.js to Special:MyPage/vector-2022.js. There are more technical details available. Interface administrators who foresee this leading to lots of technical support questions may wish to send a mass message to your community, as was done on French Wikipedia. [11]
Tech news prepared by Tech News writers and posted by bot • Contribute • Translate • Get help • Give feedback • Subscribe or unsubscribe.
MediaWiki message delivery 20:25, 22 April 2024 (UTC)
How to search under 6807 photos[edit]
I usually edit issues related to rivers of Chile and there is a lack of images directly showing that themes. Some time ago I asked here how to search for a foto of an object located between coordinates (lat1, lon1) and (lat2,lon2) under the 6807 fotos of the Category:Satellite pictures of Chile . @HyperGaruda: kindly answered here with a reference to use "Template:Object location". I tried it and became a new window on the directory page but nothing more. Can anyone give a more explizit indication?. Thanks in advance, Juan Villalobos (talk) 13:53, 28 April 2024 (UTC)
- what or how exactly do you want to find?
- try using "Map of all coordinates on OSM" i just added. RZuo (talk) 14:29, 28 April 2024 (UTC)
- structured data might allow you to search with a bounding box. Enhancing999 (talk) 14:38, 28 April 2024 (UTC)
- RZuo, the "Map of all coordinates on OSM" do it very well. I had thought in a excel file with filename;lat;lon;description. But the map is as least so good.
- Enhancing999, what is "structured data" in this case and how can I use it?
- My problem is that many regions of Chile have been never visited by anyone, not to say photografied. The last option is to show a satellite photo.
- Thank you. --Juan Villalobos (talk) 12:58, 29 April 2024 (UTC)
- if you want to search about a specific coord, here's a way to search within r km from (x,y): mw:Help:CirrusSearch#Geo_Search.
- take a look at other sections of the page for more creative ways to restrict search results.
- i dont know how to search within a box bounded by two pairs of coords though.
- Commons:Structured data. RZuo (talk) 15:02, 29 April 2024 (UTC)
- https://commons.wikimedia.org/w/index.php?search=nearcoord:99km,-38.73,-72.66 for temuco.
- https://commons.wikimedia.org/w/index.php?search=nearcoord:99km,-38.73,-72.66+deepcategory:"Satellite+pictures+of+Chile" :) --RZuo (talk) 15:18, 29 April 2024 (UTC)
- structured data might allow you to search with a bounding box. Enhancing999 (talk) 14:38, 28 April 2024 (UTC)
For your question, the following may do:
#defaultView:Map
SELECT ?file ?image ?location ?filename WHERE
{
SERVICE wikibase:box
{
?file wdt:P1259 ?location.
bd:serviceParam wikibase:cornerWest "Point(-121.872777777 36.304166666)"^^geo:wktLiteral .
bd:serviceParam wikibase:cornerEast "Point(-121.486111111 38.575277777)"^^geo:wktLiteral .
}
?file schema:url ?image;
schema:contentUrl ?url.
BIND(wikibase:decodeUri(CONCAT("File:", SUBSTR(STR(?url), 53 ))) AS ?filename)
}
LIMIT 10000
You need to pick eastmost and westmost points. Enhancing999 (talk) 23:38, 29 April 2024 (UTC)
Tech News: 2024-18[edit]
Latest tech news from the Wikimedia technical community. Please tell other users about these changes. Not all changes will affect you. Translations are available.
Recent changes
- The appearance of talk pages changed for the following wikis: Azerbaijani Wikipedia, Bengali Wikipedia, German Wikipedia, Persian Wikipedia, Hebrew Wikipedia, Hindi Wikipedia, Indonesian Wikipedia, Korean Wikipedia, Dutch Wikipedia, Portuguese Wikipedia, Romanian Wikipedia, Thai Wikipedia, Turkish Wikipedia, Ukrainian Wikipedia, Vietnamese Wikipedia. These wikis participated to a test, where 50% of users got the new design, for one year. As this test gave positive results, the new design is deployed on these wikis as the default design. It is possible to opt-out these changes in user preferences ("Show discussion activity"). The deployment will happen at all wikis in the coming weeks. [12]
- Seven new wikis have been created:
- You can now watch message groups/projects on Translatewiki.net. Initially, this feature will notify you of added or deleted messages in these groups. [20]
- Dark mode is now available on all wikis, on mobile web for logged-in users who opt into the advanced mode. This is the early release of the feature. Technical editors are invited to check for accessibility issues on wikis. See more detailed guidelines.
Problems
- Kartographer maps can use an alternative visual style without labels, by using
mapstyle="osm"
. This wasn't working in previews, creating the wrong impression that it wasn't supported. This has now been fixed. [21]
Changes later this week
- The new version of MediaWiki will be on test wikis and MediaWiki.org from 30 April. It will be on non-Wikipedia wikis and some Wikipedias from 1 May. It will be on all wikis from 2 May (calendar). [22][23]
Tech news prepared by Tech News writers and posted by bot • Contribute • Translate • Get help • Give feedback • Subscribe or unsubscribe.
MediaWiki message delivery 03:30, 30 April 2024 (UTC)
"Caught exception of type JobQueueError"[edit]
I'm trying to upload a file (DjVu, if it matters) with UploadWizard and I get an error: "Caught exception of type JobQueueError".
There's also something like [d699cdd6-f1f2-4319-a840-d9e6048a1c70], that varies.
Same with the old upload form.
Could someone explain how to fix it? Thanks, Alien333 (talk) 13:31, 4 May 2024 (UTC)
- Same here for JPG files. Here's a screenshot. Happened across different files. Re-uploaded, re-downloaded, changed a browser, and still doesn't upload. S5A-0043Talk 13:39, 4 May 2024 (UTC)
- For how long has it been doing it? Alien333 (talk) 13:41, 4 May 2024 (UTC)
- Since around half an hour ago? Can't remember exactly but I uploaded around 7-8 files including File:(SGP-Singapore) SBS Transit SBS6870S 27 2024-05-04.jpg in one go. The file I linked was the only one that passed. It wouldn't work when I tried reuploading the rest and hasn't worked since. S5A-0043Talk 13:44, 4 May 2024 (UTC)
- @S5A-0043 Now it works (at least for me). — Alien333 (what I did & why I did it wrong) 14:22, 4 May 2024 (UTC)
- Same for me here. Seems like problem solved. Also your signature made me click to see "what I did and why I did it wrong" in this scenario, lol. S5A-0043Talk 14:26, 4 May 2024 (UTC)
- It's just my usual signature, but I didn't have it here yet because I'm mostly a Wikisourcerer, and it's only upon talking here that I realised I had not set it here — Alien333 (what I did & why I did it wrong) 14:31, 4 May 2024 (UTC)
- Same for me here. Seems like problem solved. Also your signature made me click to see "what I did and why I did it wrong" in this scenario, lol. S5A-0043Talk 14:26, 4 May 2024 (UTC)
- @S5A-0043 Now it works (at least for me). — Alien333 (what I did & why I did it wrong) 14:22, 4 May 2024 (UTC)
- Since around half an hour ago? Can't remember exactly but I uploaded around 7-8 files including File:(SGP-Singapore) SBS Transit SBS6870S 27 2024-05-04.jpg in one go. The file I linked was the only one that passed. It wouldn't work when I tried reuploading the rest and hasn't worked since. S5A-0043Talk 13:44, 4 May 2024 (UTC)
- For how long has it been doing it? Alien333 (talk) 13:41, 4 May 2024 (UTC)
Error "To avoid creating high replication lag, this transaction was aborted because the write duration (..) exceeded the 3 second limit."[edit]
I keep getting that. Enhancing999 (talk) 17:12, 5 May 2024 (UTC)
- Yes, I got that too. I believe it was a temporary glitch. Yann (talk) 17:14, 5 May 2024 (UTC)
- I've been getting the error for a while now and it seems to still be going on. --Adamant1 (talk) 17:49, 5 May 2024 (UTC)
Why does this warning message start showing up after I've clicked the publish changes button?[edit]
Hi, I having trouble editing pages. The warning message say:
"To avoid creating high replication lag, this transaction was aborted because the write duration (52.137077093124) exceeded the 3 second limit. If you are changing many items at once, try doing multiple smaller operations instead."
This issue seems to have started last week. It always happens at night. On the same page, I always have to try three or even five times before the system actually accepts my edits. It bothers me a lot, so if anybody know how to solve it then help me. Thanks.--125.230.68.137 12:06, 6 May 2024 (UTC)
- I had the same problem when trying to add a category to File:Brouwerij Verhaeghe.jpg. [78b79d1d-ee4d-4bb7-a16e-3436aff4157d] 2024-05-06 19:03:00: Fatale fout van type "Wikimedia\Rdbms\DBTransactionSizeError". 5 minutes later I tried again. No problem anymore. Wouter (talk) 19:16, 6 May 2024 (UTC)
- The problem is still happening intermittently. Nurg (talk) 01:25, 7 May 2024 (UTC)
- I have opened a call at https://phabricator.wikimedia.org/T364360. Nurg (talk) 01:50, 7 May 2024 (UTC)
- I received this kind of error, too. Seems to be a larger problem --PantheraLeo1359531 😺 (talk) 15:33, 10 May 2024 (UTC)
- I have opened a call at https://phabricator.wikimedia.org/T364360. Nurg (talk) 01:50, 7 May 2024 (UTC)
- The problem is still happening intermittently. Nurg (talk) 01:25, 7 May 2024 (UTC)
New tool for importing SteamGridDB images[edit]
Hi, just wanted to share a new tool I've made: sgdb2commons. It's a browser script that makes it easy to quickly import images from steamgriddb.com. The website is a pretty good source for high-quality video game logos, and many of those logos are too simple to be copyrighted, so this tool lets you pick images to copy to Commons. Installation and usage info can be found at User:IagoQnsi/sgdb2commons. This is a very simple first version of the script, so please let me know if you encounter any issues or have any suggestions! Best, IagoQnsi (talk) 02:31, 6 May 2024 (UTC)
- I'm not certain that this site is an appropriate source for Commons media. Most of the images it offers are unofficial fan-created logos which should not be uploaded to Commons. Omphalographer (talk) 04:20, 7 May 2024 (UTC)
- It has both official and custom logos, and each logo is tagged with which type it is (albeit not always accurately). It definitely requires discretion from the uploader to determine which files are worth transferring, but I would estimate there are thousands of images on the site that would be valuable and appropriate for Commons. –IagoQnsi (talk) 04:38, 7 May 2024 (UTC)
Tech News: 2024-19[edit]
Latest tech news from the Wikimedia technical community. Please tell other users about these changes. Not all changes will affect you. Translations are available.
Recent changes
- The appearance of talk pages changed for all wikis, except for Commons, Wikidata and most Wikipedias (a few have already received this design change). You can read the detail of the changes on Diff. It is possible to opt-out these changes in user preferences ("Show discussion activity"). The deployment will happen at remaining wikis in the coming weeks. [24][25]
- Interface admins now have greater control over the styling of article components on mobile with the introduction of the
SiteAdminHelper
. More information on how styles can be disabled can be found at the extension's page. [26] - Wikimedia Enterprise has added article body sections in JSON format and a curated short description field to the existing parsed Infobox. This expansion to the API is also available via Wikimedia Cloud Services. [27]
Changes later this week
- The new version of MediaWiki will be on test wikis and MediaWiki.org from 7 May. It will be on non-Wikipedia wikis and some Wikipedias from 8 May. It will be on all wikis from 9 May (calendar). [28][29]
- When you look at the Special:Log page, the first view is labelled "All public logs", but it only shows some logs. This label will now say "Main public logs". [30]
Future changes
- A new service will be built to replace Extension:Graph. Details can be found in the latest update regarding this extension.
- Starting May 21, English Wikipedia and German Wikipedia will get the possibility to activate "Add a link". This is part of the progressive deployment of this tool to all Wikipedias. These communities can activate and configure the feature locally. [31]
Tech news prepared by Tech News writers and posted by bot • Contribute • Translate • Get help • Give feedback • Subscribe or unsubscribe.
MediaWiki message delivery 16:41, 6 May 2024 (UTC)
Suddenly the files I uploaded started to get to this category. I am sure I am not doing anything different from what I usually do (I uploaded 25+ files today, and until 15 minutes ago everything was going ok). Ymblanter (talk) 19:45, 8 May 2024 (UTC)
- @Multichill: Ymblanter (talk) 19:46, 8 May 2024 (UTC)
- I see, the issue is somewhere with the syntax, instead of hidden categories redlinks are created. Ymblanter (talk) 19:51, 8 May 2024 (UTC)
Tech News: 2024-20[edit]
Latest tech news from the Wikimedia technical community. Please tell other users about these changes. Not all changes will affect you. Translations are available.
Recent changes
- On Wikisource there is a special page listing pages of works without corresponding scan images. Now you can use the new magic word
__EXPECTWITHOUTSCANS__
to exclude certain pages (list of editions or translations of works) from that list. [32] - If you use the user-preference "Show preview without reloading the page", then the template-page feature "Preview page with this template" will now also work without reloading the page. [33]
- Kartographer maps can now specify an alternative text via the
alt=
attribute. This is identical in usage to thealt=
attribute in the image and gallery syntax. An exception for this feature is wikis like Wikivoyage where the miniature maps are interactive. [34] - The old Guided Tour for the "New Filters for Edit Review" feature has been removed. It was created in 2017 to show people with older accounts how the interface had changed, and has now been seen by most of the intended people. [35]
Changes later this week
- The new version of MediaWiki will be on test wikis and MediaWiki.org from 14 May. It will be on non-Wikipedia wikis and some Wikipedias from 15 May. It will be on all wikis from 16 May (calendar). [36][37]
- The Special:Search results page will now use CSS flex attributes, for better accessibility, instead of a table. If you have a gadget or script that adjusts search results, you should update your script to the new HTML structure. [38]
Future changes
- In the Vector 2022 skin, main pages will be displayed at full width (like special pages). The goal is to keep the number of characters per line large enough. This is related to the coming changes to typography in Vector 2022. Learn more. [39]
- Two columns of the
pagelinks
database table (pl_namespace
andpl_title
) are being dropped soon. Users must use two columns of the newlinktarget
table instead (lt_namespace
andlt_title
). In your existing SQL queries:- Replace
JOIN pagelinks
withJOIN linktarget
andpl_
withlt_
in theON
statement - Below that add
JOIN pagelinks ON lt_id = pl_target_id
- See phab:T222224 for technical reasoning. [40][41]
- Replace
Tech news prepared by Tech News writers and posted by bot • Contribute • Translate • Get help • Give feedback • Subscribe or unsubscribe.