Open zb1439 opened 4 years ago
Actually, the data split for coco is also released in folder "data". Just change the dataset config and number of classes, everything should be good.
Hi, we have tried to fix current implementation to adapt to coco dataset, with much more than we expected to modify though. We are using 4 Titan XP to reproduce results on COCO, however we face severe out of memory problem. Could you tell us what is your configuration for both batch size and subdivision? Thanks.
Thanks for this released code and nice instructions for training on VOC datasets. I am wondering what should be modified and preprocessed for reproducing your results on COCO datasets, like the preprocessing of labels and any critical hyperparameters. Thanks and waiting for your response!