I explained my problem [here](https://stackoverflow.com/questions/61304977/improving-a-pre-trained-tensorflow-object-detection-model). Any help would be appreciated.
