How to modify the prediction pipeline of a saved model?

[AlbertoSabater]

System information

• Tensorflow version: 2.1.0
• Python 3.6
• Ubuntu 18.04

Hello! I am loading (without any problem) a Faster RCNN model trained on the EPIC-KITCHENS dataset.

I want to change the prediction pipeline by modifying the non-max-suppression step in order to perform some tests. My final output should be a set of bounding boxes and a set of scoring vectors with a size of 1*C (C is the total number of classes of my dataset). However, I don't know how to do this since I cannot access to the pipeline of this trained model.

Is there any way to do this? Maybe by building a new model with my new pipeline and loading the trained weights somehow?

Thanks in advance!

[Asadullah007]

yes this is possible. possible.