Closed zhaotiny closed 7 years ago
@zhaotiny Correct me if I'm wrong - believe the code is using the annotation files as the input, not the raw images. See the code here again:
https://github.com/kwotsin/TensorFlow-ENet/blob/master/get_class_weights.py
the images come from trainannot, which is the annotation files.
when calculating the weights using "def ENet_weighing(image_files=image_files, num_classes=12)", you feed imge_files as input. But according to my understanding, we should feed annotation_files as input since we calculate the weights based on the frequencies of each class. Am I correct?