search

keras-team / keras-cv

Industry-strength Computer Vision workflows with Keras
Other
1.01k stars 331 forks source link

How to contribute pretrained models? #2445

Open veb-101 opened 6 months ago

veb-101 commented 6 months ago

Hello, I've successfully converted MobileViT v1 (variants - XXS, XS, S) models from Pytorch to Keras 3, including their pre-trained image classification weights. Can such models be added to KerasCV? If so, can you point me to the correct contributing guide?

As part of my commitment to this project, I've already created a separate repo to push other versions of MobileViT and other models.

Please try it out over here before you make any decision: https://colab.research.google.com/github/veb-101/keras-vision/blob/main/examples/mobile_vit_v1.ipynb

Repo: keras-vision

Currently, the weights are hosted in my repository releases.

sachinprasadhs commented 6 months ago

Thank you for your interest in contribution, we don't have any documentations as of now, but you can refer our existing implementations here to create a PR https://github.com/keras-team/keras-cv/tree/master/keras_cv/src/models/backbones

veb-101 commented 6 months ago

Thanks, I'll work on porting MobileViT v2 models first and then create a PR for both versions. You may close or keep this issue open.