Open applezy8866 opened 1 year ago
Dear author: There is "For each training sequence, we randomly divide it into 2 splits at the start of each training epoch to foster more diverse data sequences" in paper, but I only found fixed average split in base_dataset.py?
The random division is done by the sampler during training. You can check https://github.com/yuantianyuan01/StreamMapNet/blob/main/plugin/datasets/samplers/group_sampler.py#L267 for more details.
Dear author: There is "For each training sequence, we randomly divide it into 2 splits at the start of each training epoch to foster more diverse data sequences" in paper, but I only found fixed average split in base_dataset.py?