search

tensorflow / models

Models and examples built with TensorFlow
Other
77.23k stars 45.75k forks source link

how to do semantic segmentation on video using saved model from tensorflow-2? #10466

Open PritiDrishtic opened 2 years ago

PritiDrishtic commented 2 years ago

Currently we have TF2 on production environment so wanted to do semantic segmentation on video using saved_model.pb Video segmentation worked for frozen graph using TF-1 but TF-2 dont support frozen graph so please provide script or any working example.

System information

frederick0329 commented 2 years ago

Do you mind providing the link to code you are referring to?

PritiDrishtic commented 2 years ago

Hi @frederick0329 I am refering jupyter notebook from models/deeplab folder here I want to run above jupyter notebook for video inference.

Also refferring another jupyter notebook where she implemented run_video function for video, but this is not working with new set of models from TF-2 here

frederick0329 commented 2 years ago

Got it. Sorry, I don't work on it. Tagging the authors.

PritiDrishtic commented 2 years ago

@jaeyounkim @YknZhu @frederick0329 is there any updates on this?

frederick0329 commented 2 years ago

Sorry @jaeyounkim and I do not work on this. It might be most efficient to reach out to the maintainers. https://github.com/tensorflow/models/tree/master/research/deeplab#contacts-maintainers

frederick0329 commented 2 years ago

In the README, they also mention this. To new and existing DeepLab users: We have released a unified codebase for dense pixel labeling tasks in TensorFlow2 at https://github.com/google-research/deeplab2. Please consider switching to the newer codebase for better support.

Hope this helps.