History

Akash Manna 5af6d8b71e minor bug fix in eval (in transforms.Resize)		2020-05-22 16:00:59 +05:30
..
datasets	add attribute code	2020-05-18 18:36:55 +01:00
models	add attribute code	2020-05-18 18:36:55 +01:00
README.md	update readme	2020-05-18 22:27:15 +01:00
default_parser.py	add attribute code	2020-05-18 18:36:55 +01:00
main.py	minor bug fix in eval (in transforms.Resize)	2020-05-22 16:00:59 +05:30
train.sh	add attribute code	2020-05-18 18:36:55 +01:00

README.md

Person Attribute Recognition

This code was developed for the experiment of person attribute recognition in Omni-Scale Feature Learning for Person Re-Identification (ICCV'19).

Download data

Download the PA-100K dataset from https://github.com/xh-liu/HydraPlus-Net, and extract the file under the folder where you store your data (say $DATASET). The folder structure should look like

$DATASET/
    pa100k/
        data/ # images
        annotation/
            annotation.mat

Train

The training command is provided in train.sh. Run bash train.sh $DATASET to start training.

Test

To test a pretrained model, add the following two arguments to train.sh: --load-weights $PATH_TO_WEIGHTS --evaluate.