Faster_rcnn_resnet101_v1_640x640_coco17_tpu-8

I am running tensorflow model on google colab. But I am not getting true results and I can't figure out what I need to do in order to improve results.

1. It predict only one bounding box per image.
2. It is not predicting right object.

What parameter do I need to change in config file for more than one bounding box.
How I can improve the results to get right placement of bounding boxes on images.