Using mask rcnn data about object_level_visual_reasoning HOT 1 OPEN

Hi Nirat,

Thanks for the interest in our work.
You can find the line by line implementation about how to retrieve the bboxes/segmentations for each detected objects per frames here.
First I am retrieving the object id, then reformatting the bounding box and finally rescaling the mask.
If you feel that the predictions are not good enough you can use this pre-trained Mask-RCNN on your own data.
I hope that these explanations could help for your work.
Best,

Fabien

from object_level_visual_reasoning.