Closed SuperDQ closed 4 years ago
Allow predicted ground truth for score function(for video accuracy). Also, the low mrcnn accuracy was due to people segmentation. By adding --allclasses, the model will predict the most confident one.
Allow predicted ground truth for score function(for video accuracy). Also, the low mrcnn accuracy was due to people segmentation. By adding --allclasses, the model will predict the most confident one.