How evaluate TF1 trained and saved as pd model on tfrecord dataset

There are:

  • TF1 trained object detection model, saved as .pd.
  • TFrecord test dataset, test.tfrecord (images, annotation boxes and labels)

How may I run evaluation for this model by test.tfrecord dataset on coco metrics in TF2?

I found how run evaluation for model’s checkpoints, but I don’t have ckpt of the model. So this way doesnt work for me.

  • Ubuntu 20.04
  • TF – 2.4
  • python 3.8

Thank you.

Source: Python-3x Questions