Posted under General
This topic has been locked.
post #1758644 was uploaded only half way.
Its size is shown as 1.14 MB, which matches the original filesize on pixiv. However, when you download the fullsize file from Danbooru, the actual filesize is 562.5 KB.
And also the image hash is the correct one, so I can't even reupload it since it's flagged as a duplicate.
Were you shooting for topic #8984?
Log said:
Were you shooting for topic #8984?
No, I was just reporting a bug. The post is already deleted so there's no action required.
"Find Similar" seems nonfunctional to me. Using it on post #1762312 didn't give me post #1426202, despite the obvious similarity. Would it be problematic to restore the display of weaker matches and their degree of similarity? Having the matched post itself excluded from the list of similar posts would also be nice.
Flopsy said:
"Find Similar" seems nonfunctional to me. Using it on post #1762312 didn't give me post #1426202, despite the obvious similarity. Would it be problematic to restore the display of weaker matches and their degree of similarity? Having the matched post itself excluded from the list of similar posts would also be nice.
That post is below the threshold of what counts as similar enough to be displayed. It only has 70% similarity, while only 80%+ posts are displayed.
It would be possible to lower the threshold so that the post would be returned. But lowering the threshold too much could cause false positives to start to be displayed, so a good balance would need to be found.
What kind of damage could a false positive reasonably do? People aren't going to go "Hey that image looks nothing like this one but the computer says it's similar so I'd better flag this one anyway.", are they? (If someone actually did, it would be an easily correctable mistake.)
If bringing back the old interface for similar images isn't an option, could you at least have it display the three best matches (not including the image itself) every time, without an absolute cutoff level?
Edit: Slight rephrasing.
Updated
Flopsy said:
What kind of damage could a false positive reasonably do?
A false positive every now and then wouldn't be the end of the world, but if there started being more false positives than correct results then the feature would not be very usable, hence why I said a "balance".
Flopsy said:
People aren't going to go "Hey that image looks nothing like this one but the computer says it's similar so I'd better flag this one anyway.", are they? (If someone actually did, it would be an easily correctable mistake.)
"Duplicate" is not a valid flag reason, even when the post actually is a duplicate, so no one should be flagging posts for that reason to begin with.
Do a Ctrl+F5 or clear your browser cache. http://en.wikipedia.org/wiki/Wikipedia:Bypass_your_cache
evazion said:
Do a Ctrl+F5 or clear your browser cache. http://en.wikipedia.org/wiki/Wikipedia:Bypass_your_cache
Hmm, nope. No method of clearing the cache seems to work. Screenshot of the error console http://i.imgur.com/HQFdCGJ.png
Looks like there were some very minor javascript changes in the last update, but nothing I see that would break everything in old browsers. Normally stuff like this is a caching problem.
So triple check that your cache is clear. Try visiting the site in incognito mode / private browsing mode (close and reopen it first if you're already in it). That should definitely bypass the browser cache.
evazion said:
Looks like there were some very minor javascript changes in the last update, but nothing I see that would break everything in old browsers. Normally stuff like this is a caching problem.
So triple check that your cache is clear. Try visiting the site in incognito mode / private browsing mode (close and reopen it first if you're already in it). That should definitely bypass the browser cache.
Nope, private browsing mode does not work.
Whelp looks like you're right, it is a bug: issue #2243.
Tonny890 said:
Trying to bring up any artist wiki page gets me "Something Broke"
More of issue #2234, I guess.
