resume training

Hi, I download pretrain model and created lmdb of my train and validation dataset. I divide my dataset for training(50%), testing(30%) and validation(20%). my question is that how to modify train command : --train_nips path_to_dataset \ --train_cvpr path_to_dataset \ --valroot path_to_dataset \

while train_nips and train_cvpr are required but i have one dataset. thanks.

[Canjie-Luo]

An easy way is to set the train_nips and train_cvpr the same. Or modify the https://github.com/Canjie-Luo/MORAN_v2/blob/2cd40c41d6041f54efdbdcb7d7e081231168c468/main.py#L75 to train_dataset = dataset.lmdbDataset(root=opt.train_root, transform=dataset.resizeNormalize((opt.imgW, opt.imgH)), reverse=opt.BidirDecoder) You should modify all related code in main.py.

[sssimpleboy]

是的，我将那两个设置成一样的路径了。

大佬，他有三个路径，我只有一个数据集，我把前两个设置成一样，那么第三个该怎么办呢？谢谢