training without camera id

[BabuTango]

@layumi I have images from mulitple camera, but I don't have the camera information. Can I train with out specifying camera id's.

Thanks

[layumi]

Hi @BabuTango The training process does not need camera id. We only use the camera ID when evaluating.

[BabuTango]

Thanks @layumi . Should we write our own class for training on custom dataset

[layumi]

I think you only need to modify the prepare.py to put the images of the same person into each folder. You still could use the train.py to train a model.

[bmiftah]

@layumi In the prepare_viber.py - it looks it prepare the data in a different way than the other prepare.py for market and duke - By this I mean , it generate 6 test folder each with gallery and query folder .. DO these test folder refers to different splits ? I was trying to use this dataset to user it in train.py but some directories like /train (like bounding_box_train in duke and market) is missing to prepare train and val as expected by the dataloader of train.py .. . gallery and train prepared in /test ..but for train ... which one to use ? --- From -- > Viber/cam_a/all , Vibercam_b/all put them together in /train_all as in prepare.py ( I may have the wrong assumption )