google-ai-edge / mediapipe

Cross-platform, customizable ML solutions for live and streaming media.
https://ai.google.dev/edge/mediapipe
Apache License 2.0
27.4k stars 5.15k forks source link

MobileNetV4 support in Model Maker #5538

Open Tachi107 opened 3 months ago

Tachi107 commented 3 months ago

MediaPipe Solution (you are using)

Model Maker

Programming language

No response

Are you willing to contribute it

No

Describe the feature and the current behaviour/state

MobileNetV4 is the latest generation of MobileNets, featuring numerous improvements. It'd be nice to add the possibility of using this new generation instead, since Model Maker is currently stuck at MobileNetV2

Will this change the current API? How?

No response

Who will benefit with this feature?

All users of Model Maker for image classification and object detection

Please specify the use cases for this feature

Improvement on an existing feature

Any Other info

No response

joezoug commented 2 months ago

HI @Tachi107,

Thanks for the suggestion. Unfortunately the team currently doesn't have bandwidth to update the MobileNet architecture in the near future.

If there is a tfhub or kaggle checkpoint of MobileNetV4, then you can try to add it in to the uri of the ModelSpec for your use cases.

github-actions[bot] commented 2 months ago

This issue has been marked stale because it has no recent activity since 7 days. It will be closed if no further activity occurs. Thank you.

Tachi107 commented 2 months ago

This isn't a bug, but a feature request. Therefore it should be kept open to track the feature request, or closed as wontfix. There's no reason to mark this as "stale".