One class custom dataset
Hello, I am using a dataset that has only one class. I have modified the .yml file and set num_classes to 1, but I get errors. When I set though num_classes to 2 the training begins as expected. Can you provide some info on what to do with datasets that have 1 class?
Hello, I am using a dataset that has only one class. I have modified the .yml file and set num_classes to 1, but I get errors. When I set though num_classes to 2 the training begins as expected. Can you provide some info on what to do with datasets that have 1 class?
an additional class might be background
Hello, I am using a dataset that has only one class. I have modified the .yml file and set num_classes to 1, but I get errors. When I set though num_classes to 2 the training begins as expected. Can you provide some info on what to do with datasets that have 1 class?
an additional class might be background
So you would suggest when dataset has 1 class to set num_classes to 2?
Hello, I am using a dataset that has only one class. I have modified the .yml file and set num_classes to 1, but I get errors. When I set though num_classes to 2 the training begins as expected. Can you provide some info on what to do with datasets that have 1 class?
an additional class might be background
So you would suggest when dataset has 1 class to set num_classes to 2?
yes
@ShihuaHuang95 could you confirm?
Hello, I am using a dataset that has only one class. I have modified the .yml file and set num_classes to 1, but I get errors. When I set though num_classes to 2 the training begins as expected. Can you provide some info on what to do with datasets that have 1 class?
me too
same issue, can someone give the explaination? THANKS!
Hello, I am using a dataset that has only one class. I have modified the .yml file and set num_classes to 1, but I get errors. When I set though num_classes to 2 the training begins as expected. Can you provide some info on what to do with datasets that have 1 class?
The bug still exists, and I hope the authors can fix it soon. The robustness of academic code is still not as reliable as that of commercial closed-source code.