search

matterport / Mask_RCNN

Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow
Other
24.52k stars 11.68k forks source link

How to create dataset for two or more classes like balloon dataset...? #780

Open matiqul opened 6 years ago

jlognn commented 6 years ago

Please provide some more information. Why don't you just use the approach given in balloon.py?

matiqul commented 6 years ago

I would like to train my own dataset with many class....

cardboardcode commented 6 years ago

Have a look at the following project based on Mask_RCNN: https://github.com/SUYEgit/Surgery-Robot-Detection-Segmentation

Specifically, line 91 under surgery.py of the aforementioned repo, it serves as a good reference on how you can start adding additional classes of object to detect in the original coco.py file.

As to how to annotate the training input images to suit the extended code, please refer to line 107 of surgery.py for the exact .json format needed. Try comparing this file with the balloon.py under the balloon color splash tutorial for a clearer idea.