Fine-tuning Tensorflow 2 Model Zoo models?

[TensorOverflow]

Hi,

Can someone from the Tensorflow Team please provide information on how to fine-tune a model from the Tensorflow 2 Model Zoo when training it for object detection?
I have read that there is a freeze_variables parameter in the pipeline.config file, but is this still usable?
Let’s say you want to freeze all the layers, expect the top 10, how can this be done when doing object detection?

Thanks!

[Bhack]

You could check Adding community contributed guides by sayakpaul · Pull Request #9271 · tensorflow/models · GitHub

[lgusm]

On the OD repo there’s a lot of useful information that might help you:

• list of colabs with code: models/research/object_detection/colab_tutorials at master · tensorflow/models · GitHub
• Few shot code - models/eager_few_shot_od_training_tf2_colab.ipynb at master · tensorflow/models · GitHub