azeaa

I made a data set similar to s3dis, with three classes marked, but the background was not marked. After training the backbone network, it predicts the background as the first...

Can I train on my own dataset? What is the format of the dataset?