Hello, I am a student specializing in deep learning. What confuses me is that the generation of IOU predictions does not involve the use of activation functions like sigmoid or softmax. How can we ensure that the output results are between zero and one? Additionally, how do you calculate the loss between IOU predictions and the actual IOU when fine-tuning your model? Thank you.
Hello, I am a student specializing in deep learning. What confuses me is that the generation of IOU predictions does not involve the use of activation functions like sigmoid or softmax. How can we ensure that the output results are between zero and one? Additionally, how do you calculate the loss between IOU predictions and the actual IOU when fine-tuning your model? Thank you.