Closed wwjwy closed 2 years ago
In addition, I print the confmatrixs, but i am confused which is correct as show below. the left or the right is the correct confmatrixs? "conf_mat.update(pred.argmax(1).flatten(), test_label.flatten())"
All results are reported from validation set averaged across three random seeds. This is the standard in semi-supervised segmentation.
For your second question, it should be the second graph.. but it doesn't really matter as long as you keep consistent with the choice of the dimension.
Thanks a lot!
Hi, In your voc dataset split, I noticed that there are only training and val set. Are there additional test sets to verify model performance after you have completed training 200 epochs? The voc result in Table 1 in the paper is on the validation set or the test set? Thanks!