search

chufengt / ALM-pedestrian-attribute

Code for the paper "Improving Pedestrian Attribute Recognition With Weakly-Supervised Multi-Scale Attribute-Specific Localization", ICCV 2019, http://arxiv.org/abs/1910.04562.
Apache License 2.0
188 stars 66 forks source link

How have you made the datasets ready? #4

Closed Ehsan-Yaghoubi closed 4 years ago

Ehsan-Yaghoubi commented 4 years ago

Could you please explain about the dataset preparation. Is it similar to this project? (https://github.com/dangweili/pedestrian-attribute-recognition-pytorch)

I used this script (https://github.com/dangweili/pedestrian-attribute-recognition-pytorch/blob/master/script/dataset/transform_peta.py) to prepare the peta dataset. It is producing a .pkl file. But, in your codes, you don't analyze .pkl files!

What should be the type of the label in the __init__ function of the MultiLabelDataset class?

I appreciate your clarification about dataset preparation.

chufengt commented 4 years ago

Yes, it's not exactly the same as dangweili's project, but they provide a good example.

The label in MultiLabelDataset class means the file that encoding the mapping between image name and its ground-truth (GT vector should contain only 0 or 1 values).

For PETA and PA-100K, they provide the same format, just download it from the official websites (be careful about GT values since some of them using -1 or 1). For RAP, they only provide .mat files, which need to be processed according to their ReadMe.txt (sorry I can't find my scripts but it's not difficult).

chufengt commented 3 years ago

I've uploaded the label lists.

eeric commented 3 years ago

Could you please explain about the dataset preparation. Is it similar to this project? (https://github.com/dangweili/pedestrian-attribute-recognition-pytorch)

I used this script (https://github.com/dangweili/pedestrian-attribute-recognition-pytorch/blob/master/script/dataset/transform_peta.py) to prepare the peta dataset. It is producing a .pkl file. But, in your codes, you don't analyze .pkl files!

What should be the type of the label in the __init__ function of the MultiLabelDataset class?

I appreciate your clarification about dataset preparation.

https://github.com/chufengt/iccv19_attribute/issues/37

eeric commented 3 years ago

https://github.com/chufengt/iccv19_attribute/issues/37