Question about Training and Testing Sets
Yangxc992 opened this issue · comments
Thank you for your interesting work!
I would like to know how many images were used for training and testing on the Cityscapes dataset. Are the training and testing sets the same? I have read your paper carefully, but this information is not mentioned.
Not same. For Cityscape dataset, there are 5000 images at 1024 * 2048 resolution pre-split into training (2975), validation (500), and test (1525) sets. We set the bit error rate = 10^-5 for both training and testing sets. Since the bit errors happen randomly, the caused error pattern is unpredictable. We use the training sets (2975) for our network training and evaluate the model in the testing set (1525).
Not same. For Cityscape dataset, there are 5000 images at 1024 * 2048 resolution pre-split into training (2975), validation (500), and test (1525) sets. We set the bit error rate = 10^-5 for both training and testing sets. Since the bit errors happen randomly, the caused error pattern is unpredictable. We use the training sets (2975) for our network training and evaluate the model in the testing set (1525).
Thank you for your reply!