KVMM (Keras Vision Models) : "Timm" for Keras 3!

[IMvision12]

Keras Vision Models (KVMM), a comprehensive collection of vision models with pre-trained weights entirely in Keras 3. KVMM supports a wide range of computer vision tasks including segmentation, classification, and more to come (object detection and vision-language modeling).

✨ Key features:

• 25+ backbone architectures with various pre-trained weights (Swin, ViT, ResNeXt etc)
• Support for multiple weight variants (in1k, in21k, augreg_in1k, and many more)
• Build segmentation models (Segformer) with any custom backbone within KVMM
• Custom layers and backbone flexibility
• Full Keras 3 compatibility

Link : https://github.com/IMvision12/keras-vision-models