Evaluate after training for a step or epoch for object detection

[DevanshBheda]

Prerequisites

Please answer the following question for yourself before submitting an issue.

• [Y] I checked to make sure that this feature has not been requested already.

1. The entire URL of the file you are using

https://github.com/tensorflow/models/blob/master/research/object_detection/model_main_tf2.py

2. Describe the feature you request

If we could evaluate across eval data while training, that would be really helpful. The same feature was present in tf1 version of object detection api.

3. Additional context

There is an option to continuously evaluate our checkpoints. But that requires a GPU and takes resources away from training process. If we could merge them into a single process, that would be helpful.

4. Are you willing to contribute it? (Yes or No)

Yes

[aavishkarmishra]

Hello @tombstone @jch1 @pkulzc , I would like to contribute to this issue.