Closed pshivraj closed 5 years ago
This PR add further improvements to the training pipeline:
Model - Added two new methods to train just specific masks.
mrcnn_mask: Just mask layers mask_heads: Mask layers or rpn/fpn
Utils - Added weighted average to AP calculation with updated mAP calculation range for IoU being 0.5-0.8
Train - Updated the mask load function to avoid difference in number of masks and class id's, additionally current deep augmentations being used.
pre_process - Updated the mask load function to avoid difference in number of masks and class id's
Inference_notebook - This notebook contains the visualization to analyze results. #35
Finally this PR also adds the Test data set to the repo.
Let us know when this is ready to review/merge 😄
I am done with the current PR, you can review/merge it.
This PR add further improvements to the training pipeline:
Model - Added two new methods to train just specific masks.
Utils - Added weighted average to AP calculation with updated mAP calculation range for IoU being 0.5-0.8
Train - Updated the mask load function to avoid difference in number of masks and class id's, additionally current deep augmentations being used.
pre_process - Updated the mask load function to avoid difference in number of masks and class id's
Inference_notebook - This notebook contains the visualization to analyze results. #35
Finally this PR also adds the Test data set to the repo.