Commons talk:WMF support for Commons/Upload Wizard Improvements/Logo detection
What is the problem we are trying to solve here?
[edit]What is the problem we are trying to solve here? Clearly, a human can look at these and easily determine whether they are logos or not. What advantage does a machine bring to this? Is this just to say that if these aren't in logo categories they should be? Or what? I take it that this is probably about finding copyvios, so let's look at what it found in those terms.
Looking at the examples on this page that are 98% or higher, as of this writing we have
- Already correctly marked as below the threshold of originality (4):
- PD work of a government (1):
- Dubious authorship/licensing claim, but probably (in some cases certainly) below the threshold of originality (10):
- Dubious authorship/licensing claim, very close to the threshold of originality (4):
- File:EquityZen Logo.png
- File:STEG STE TUNISIENNE DE L'ELECTRICITE ET DU GAZ.png
- File:Mactealogo1999.png (currently nominated for deletion, maybe nominator found through this page)
- File:Ifcem logo main reg.png
- Dubious authorship/licensing claim, over the threshold of originality, but might have enough historical precedent to be uncopyrightable (1):
- Dubious authorship/licensing claim, probably over the threshold of originality, probably should be deleted (2)
- File:Konnoznet Logo.png (currently nominated for deletion, maybe nominator found through this page)
- File:The Shambles Logo.jpg
So out of 22 "positives"
- 5 correctly marked as no problem
- 10 that are ultimately no problem, but probably have the wrong licensing claim
- 5 that may have a problem, and have the wrong licensing claim
- 2 that look to me like pretty clear deletes.
To me, the most interesting result here is the high percentage that simply need their tag fixed ({{Textlogo}} rather than a dubious licensing claim). It suggests to me that we should somehow make that possibility easier to find at upload time.
Also: if we had a page that would report the result of the 98%+ "hits" here with current licensing, source, and authorship claims so that the reader did not need to click in, that might be a useful resource. It would be particularly useful to see logos with claims of "own work" and a CC license, which are almost never valid. - Jmabel ! talk 14:49, 9 April 2024 (UTC)
- @Jmabel: I tagged File:The Shambles Logo.jpg {{subst:npd}}. — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 15:08, 10 April 2024 (UTC)
- @Jmabel Thanks for your analysis (and sorry for the late reply, but for some reasons I lost your intervention). Right now, we're exploring with the community how this logo detection tool can be integrated in your workflow, so all suggestions are welcome. The idea of tagging logos claiming to be "own work" is actually good, and we can work on that. Sannita (WMF) (talk) 13:58, 11 April 2024 (UTC)
- @Sannita (WMF): Can the tool automatically tag own work logo uploads by new users as {{subst:npd}}? — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 14:02, 11 April 2024 (UTC)
- I also suggested to separate the "Own work" claimed logos from the rest in my suggestion in the technical village pump. Oftentimes, PD-textlogo images are tagged as "Own work" and given a restrictive copyright ©️ license when in reality they were never eligible for copyright restrictions to begin with, it's just that many novice uploaders are unfamiliar with how to properly tag and organise them. Tagging them with a "No licence" template makes them eligible for automated deletion without a human reviewer ever having laid their eyes upon them and I'd rather have the human reviewers tag them as they are more likely more familiar with the intellectual property laws of the country of origin of the logo and can determine if it is or isn't above the threshold of originality or the jurisdiction concerned. --Donald Trung 『徵國單』 (No Fake News 💬) (WikiProject Numismatics 💴) (Articles 📚) 14:07, 11 April 2024 (UTC)
- @Jeff G.: I don't think {{Npd}} is what we want here: most of these will end up as {{Textlogo}}. Probably use a distinct tag, because we will want these reviewed by people who know copyright. And not a lot of point in pinging the uploader for that, because they are probably clueless about TOO. - Jmabel ! talk 17:54, 11 April 2024 (UTC)
- @Jmabel: Fine, a new template/tag, then. What would you want to call it? {{Logo detected}}, categorizing to Category:Logos to be checked in Category:Logos? — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 00:41, 12 April 2024 (UTC)
- That's fine, though the problem would be better characterized by calling the template {{Logo + own work detected}}. - Jmabel ! talk 00:45, 12 April 2024 (UTC)
- @Jmabel: Fine, a new template/tag, then. What would you want to call it? {{Logo detected}}, categorizing to Category:Logos to be checked in Category:Logos? — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 00:41, 12 April 2024 (UTC)
- @Sannita (WMF): Can the tool automatically tag own work logo uploads by new users as {{subst:npd}}? — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 14:02, 11 April 2024 (UTC)
New designs for logo detection tool
[edit]Hello all! We're happy to share that we will work on logo detection in the following months and that we defined an initial approach for this.
After careful consideration of all the pros/cons of explored concepts, for now the tool will only alert users that the image might be a logo and that, if it doesn't meet the guidelines, it might be deleted. This approach will be the least intrusive possible, and will not add more steps to upload, to prevent extra-clicks. It will work also in case of multiple uploads, by having the warning close to the affected image.
We would like to have your feedback about it: what do you think of our initial mockup? Would it work in your opinion?
We also want to know which categories of images, in your opinion, should be excluded from the current approach (basically to avoid false positives). Is there any category that might be problematic? We already know about coat of arms and flags that have CoA in them, are there more examples?
Let us know! Sannita (WMF) (talk) 13:51, 11 June 2024 (UTC)
- The approach has great potential. Hopefully eventually it will work for other fields as well.
- I understand that you might not want to add steps at this stage. Maybe the message could list a few plausible alternatives (even if the uploader can't click them directly in the current version). Also, at least an automated category could be added for user review. Ideally with the possibility to customize both message and categorization depending on the percentage (maybe we don't need the category if it's 99.9% or it should be a different one). Possibly the message could be deactivated and only the category added (e.g. <70%), etc. One could image that the detection rules are also customizable in an interface message and allow additional groups of rules to be added. About what's ideal for logos, the logo specialists will want to weight-in. Enhancing999 (talk) 22:07, 11 June 2024 (UTC)
- I like the idea, but think that the alert should link to a page that has all the relevant info, currently this is spread out across 4+ pages, and should lead to an article that summarizes everything applicable. This will probably need to be something written by WMF legal and permanently protected, due to the high visibility. I like where this is going, but definitely needs to have anything with 50% certainty or higher get a hidden cat marking it for review, and 90% and above a separate cat, with a template added until reviewed. If license review wasn't critically understaffed, I would have them do it, but right now, I think that it should be done by patrollers (only 144 people have logged patrol actions on commons in the last 30 days, and most of them either aren't active in CVN work, or are admins/license reviewers).
- All the best -- Chuck Talk 23:07, 11 June 2024 (UTC)
- @Enhancing999 @Alachuckthebuck Thanks for your comments. We'll take into consideration your suggestions. I'm a bit wary of Legal drafting a Commons policy though: it would be better if the community would set it up initially, because in our opinion it's community that knows better what are the needs of the project. As for the implementation, as I said, we're currently aiming for something not intrusive at all, but with enough consensus with the community we can change that. Our only worry is to not add more work on admins/patrollers, at least not without their explicit consensus. This is why we settled on a simple alert. Thanks again for your comments! Sannita (WMF) (talk) 10:41, 13 June 2024 (UTC)
- If so, how would anybody (but the uploader) even know about it? Enhancing999 (talk) 10:44, 13 June 2024 (UTC)
- @Enhancing999 Apologies for answering late, I just lost the comment in my notifications. For the moment, we are counting on a "chilling effect" given by the notice. Then, if we got more consensus on how to implement different strategies than this one, we can easily adapt to what is community consensus. Sannita (WMF) (talk) 12:26, 14 June 2024 (UTC)
- If so, how would anybody (but the uploader) even know about it? Enhancing999 (talk) 10:44, 13 June 2024 (UTC)
- @Enhancing999 @Alachuckthebuck Thanks for your comments. We'll take into consideration your suggestions. I'm a bit wary of Legal drafting a Commons policy though: it would be better if the community would set it up initially, because in our opinion it's community that knows better what are the needs of the project. As for the implementation, as I said, we're currently aiming for something not intrusive at all, but with enough consensus with the community we can change that. Our only worry is to not add more work on admins/patrollers, at least not without their explicit consensus. This is why we settled on a simple alert. Thanks again for your comments! Sannita (WMF) (talk) 10:41, 13 June 2024 (UTC)
- As marketers always upload logos, perhaps the message could include reminder like "You can follow the Volunteer Response Team process to release permission if you or your employer own the copyright of the logo"? Thanks. SCP-2000 12:03, 25 June 2024 (UTC)