Closed wuwuwuxxx closed 2 years ago
Benchmarking data (e.g. images part of the benchmark testset) are not to be used until inference for leaderboard submissions; not for parameter tweeking and not for any training. If an unrelated public dataset offers a second validation set with public GT which they call "test data", then this may be used during training as well.
hi, for obj. segmentation during training can we use test data for some unsupervised learning method?