In case anybody was wondering about how I monitor progress in DeOldify- I don’t pay a huge amount of attention to training/validation loss because they don’t actually capture a lot of the issues I wind up seeing (artifacts, uniformly black renders, etc).https://twitter.com/citnaj/status/1200854194002878464 …
-
Show this thread
-
And then I have both the Jupyter notebook “test” image suites and then a batch of 100s of more public domain images. It’s painful and tedious but using my eyes on these directly I’ve found is the only reliable way I have to measure progress. And even the you have to be extremely
1 reply 0 retweets 10 likesShow this thread -
careful about perception issues- you misremember stuff easily, and literally see things differently depending on context. So I don’t necessarily take a first go at it as the final say as to whether or not the model is great or not.
2 replies 0 retweets 5 likesShow this thread -
Replying to @citnaj
Your feedback to what is good, and what is great, eg. the human perceptual things you ended up focusing on, would be very interesting! I imagine trying to approximate your process with a loss function would lead to improvement in other domains.
1 reply 0 retweets 1 like
Translating “it looks cool” to something tangible and repeatable like that is tough! But yeah it would really be beneficial.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.