Optimize NMS usage by passing in an argument to the MaskRCNN and FasterRCNN constructors (they should have the right option) to set the NMS threshold. Then, remove the usage of our NMS call in the ML pipeline (but keep the function definition around for future usage).
What's currently happening is there's 1 NMS that's run within the constructor, and 1 that we run in our pipeline, which is redundant.
Optimize NMS usage by passing in an argument to the MaskRCNN and FasterRCNN constructors (they should have the right option) to set the NMS threshold. Then, remove the usage of our NMS call in the ML pipeline (but keep the function definition around for future usage).
What's currently happening is there's 1 NMS that's run within the constructor, and 1 that we run in our pipeline, which is redundant.