To utilise already labelled RGB images with polygon annotations for training SD-maskrcnn

BerkeleyAutomation / sd-maskrcnn

Code for SD Mask R-CNN Project

https://sites.google.com/view/wisdom-dataset/home

Other

213 stars 53 forks source link

To utilise already labelled RGB images with polygon annotations for training SD-maskrcnn #61

Open MadhuriPatil1694 opened 2 years ago

MadhuriPatil1694 commented 2 years ago

I have a JSON file(coco format) that contains the image path, name, size, bbox, mask in bitmask format. I also have the corresponding depth image. But I am a little bit confused about how can I convert it into dataset format for this task. Can someone please guide me?

MadhuriPatil1694 commented 2 years ago

Should I make a different coco format JSON file for the depth images with the same mask co-ordinates