Kaiyang

See https://github.com/KaiyangZhou/deep-person-reid/blob/master/LICENSE

Yes, you're right. Some data might not be available for commercial use, so the pre-trained models are supposed to be used for research purpose only.

looks like a mis-registration error if you're building a new trainer, please do like this https://github.com/KaiyangZhou/CoOp/blob/main/trainers/coop.py#L211 to register the trainer

https://github.com/KaiyangZhou/CoOp/blob/main/datasets/oxford_pets.py#L11

https://github.com/KaiyangZhou/CoOp/tree/main/lpclip

follow this https://github.com/KaiyangZhou/Dassl.pytorch#add-a-dataset to write a dataset class

https://github.com/KaiyangZhou/Dassl.pytorch#add-a-dataset

There is a config variable controlling how many times you apply transformation to an image: https://github.com/KaiyangZhou/Dassl.pytorch/blob/master/dassl/config/defaults.py#L96 You can simply set this variable to 2 and the data loader will return...

Is 95% the reported number? If so, have you checked whether your param setting is the same as the paper? For the sgd optimizer, I'd suggest using a larger `LR`,...

it might be possible: only apply the collaborative learning to the shared labels (you need to double-check if the current implementation supports this)