search

theophanemayaud / video-simili-duplicate-cleaner

https://theophanemayaud.github.io/video-simili-duplicate-cleaner/
GNU General Public License v3.0
51 stars 5 forks source link

Make this into a library or CLI app #111

Open LifeIsStrange opened 1 year ago

LifeIsStrange commented 1 year ago

Being a library or command line app would enable software to automatically use this to find duplicates. If only a GUI is provided then this require manual work and cannot be fed into e.g a web crawler pipeline. This is important has there are many use cases @theophanemayaud friendly ping

Also note that neural networks can be complementary to an SSIM/metric, e.g. https://paperswithcode.com/sota/image-similarity-detection-on-disc21-dev or https://github.com/facebookresearch/vsc2022

besides neural networks, there are newer image metrics that better capture human perception, such as https://github.com/google/butteraugli and the newer https://github.com/cloudinary/ssimulacra2 which is a superset of SSIM Both are extensively used for tracking progress of the successor of the JPEG codec, JPEG XL

theophanemayaud commented 1 year ago

The CLI interface would be great ! It’s a totally new way to interact with the app so I don’t know where to start, though. I’m not sure I will be able to do something good soon.

The other algorithms for similarity scans are great ideas, I think the current ones are quite old. I’d love to also do some local retraining, with a nn that fine tunes on what the user manually selects as matches. I’m really interested in this, thanks for the starting tips !