Six Tiny AI Models That Run Entirely in Browser or on CPU, With Source Codes
Thanks for the love on my previous articles. Everyone wants the source code, and I can't send them one by one, so I've organized everything below for easy download.
Also, some friends specifically requested the server-side background removal (SAM3.1, 3.21GB file). That article is done too, I'll check it and publish this week. Stay tuned.
- Background Removal
https://pan.baidu.com/s/1IT34ypSHtZzvljNpztm7kg?pwd=tcmh
- OCR Text Recognition
Web-based OCR, loads a 6MB model, fast and accurate, Baidu is really great this time
https://pan.baidu.com/s/1x5SLZc8tf630DkiFbU9ITQ?pwd=urkj
- Restoring Old Photos
https://pan.baidu.com/s/1ERg0OZosUwlUHrice7kvlg?pwd=w8uf
- Torso Recognition
Web-based, 5.5MB Google model, recognizes torso, video doesn't lag
https://pan.baidu.com/s/1mYligk3qvVSWlBS\_BnleSQ?pwd=gr6w
- Face Recognition
Web-based, 6.5MB face recognition model, Google framework, fast and accurate
https://pan.baidu.com/s/1l\_g6OtN6b8K\_ZxGAm9r\_yA?pwd=n12h
- Speech to Text
Alibaba ASR speech-to-text, using JS to simulate WeChat
https://pan.baidu.com/s/13v2abmQ7NxJJCdvcrokZEg?pwd=pqfk
- To Be Continued
I'll write more articles about small models later and release collections regularly. Thanks for reading.
Top 1 from juejin.cn, machine-translated. The original thread is authoritative.
Where's the source code?