How to Train for Custom dataset

[FatimaZulfiqar]

Hi. Will you please tell me how can I prepare and train for custom dataset like Market-1501?

[Yukun-Huang]

Reorganize your own data following the file structure of the LIP dataset.

[FatimaZulfiqar]

Yes but I have downloaded the LIP dataset but I am unable to reorganize Market-1501 dataset. The file structure for both datasets are pretty different. Can you provide me with the script or is there any way to do this? Actually I am trying to create part masks on Market-1501 dataset for Person ReID task

[Yukun-Huang]

Human pasing and ReID are two different tasks. You cannot train on Market-1501 because there are no human pasing labels.
Directly using the model weights trained on the LIP dataset is a feasible solution.

[FatimaZulfiqar]

ohhh okay thankyou. But I have one more confusion that can weights of model that was originally trained on LIP dataset be tested and evaluated on market-1501 dataset directly? or do I also have to reorganize Market-1501 for testing also?

[Yukun-Huang]

Yes, the model trained on LIP can generalized to Market. No need to rereganize Market-1501, directly using inference.py to extract the mask is ok.

[FatimaZulfiqar]

ok thank you so much

[FatimaZulfiqar]

Hi I am getting following errors. I was hoping maybe you can help me to fix it

[SuperMamba]

Hello, the data set link is invalid.