search

Event-AHU / OpenPAR

[OpenPAR] An open-source framework for Pedestrian Attribute Recognition, based on PyTorch
MIT License
63 stars 6 forks source link

How to train PromptPAR using our own dataset #24

Closed VisImage closed 2 weeks ago

VisImage commented 2 months ago

Thank you for sharing the code, but how to train PromptPAR using our own dataset?

1125178969 commented 2 months ago

You can refer to the code in /dataset/prepocess to generate pkl files of your own dataset for training. Just include the image path and onehot labels

VisImage commented 2 months ago

Thanks for reply. There are more than 10 files in the folder. Not clear which file (or files) I need to read to get our own dataset?

1125178969 commented 2 months ago

You can refer to pa100k.py because he is simple enough, you just need to modify attr_words, pa100k_data (annotation file) to your own dataset.

VisImage commented 2 months ago

Thank for the reply. but I am not able to find the file pa100k.py in the folder /dataset/preprocess as you indicated. the files in the folder are listed below. Please let me know if that is the right folder. Thank you

OpenPAR_VTFPR++_dataset_preprocess idea

1125178969 commented 2 months ago

you can find it in PromptPAR