Exporting a trained model for inference (Tensorflow 2)

[MattVoge]

I have a general question: How can I export a Object Detection model including all weights into a protobuf file without using checkpoints? Tensorflow 1 allowed this (export_inference_graph.py), however, I see no way in TF2.

[lucky-xu-1994]

I have the same question with @MattVoge .

[saikumarchalla]

@MattVoge @lucky-xu-1994 Could you please use exporter_main_v2.py for TF2 and let us know if that solves your query. Thanks!

[lucky-xu-1994]

@MattVoge @lucky-xu-1994 Could you please use exporter_main_v2.py for TF2 and let us know if that solves your query. Thanks!

I have used exporter_main_v2.py to export model, it created a saved_model (assets, variables, saved_model.pb),and can only be used as an inference model, not supported to be used on other platforms.
so what I need is to freeze the model like in TF1, please help!

[MattVoge]

@MattVoge @lucky-xu-1994 Could you please use exporter_main_v2.py for TF2 and let us know if that solves your query. Thanks!

Thanks for your reply. But as I've said above, I want to export it without checkpoints. Just one protobuf including all weights and stuff.

[romeolandry]

I have the same Problem. Please Help!