Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

This repository re-implements the ECCV 2018 paper Deep Imbalanced Attribute Classification using Visual Attention Aggregation

Development Environment

  • Python 3.5

  • MXNet with CUDA-9

$ pip install --upgrade mxnet-cu90
  • Add project path to PYTHONPATH
$ export PYTHONPATH=/project/path:$PYTHONPATH
$ cd /project/path

Download Datasets

  • WIDER-Attribute: The original images and the annotation files are provided here, cropped images for each human bounding box can be downloaded here. 28,340 cropped images in "train" and "val" for training, 29,177 cropped images in "test" for testing.

Prepare Data

All records, list and txt files should be provided in the wider_records/ folder

Pre-trained Models

Pre-trained models should be placed in the saved_models/. Comment-in the lines that fetch pre-trained models the first time you train it. Then save the corresponding models to the folder.


  • Place the WIDER-Attribute dataset under the path /dataset/path/WIDER/. Then copy paste the images and rename as before to Image_cropped/.

  • Call the resize_images function from preprocessing/ to resize all images to 256x256 and save them.

  • Place the downloaded annotation text files under /dataset/path/WIDER/wider_att/.

  • Call the data_prep function from preprocessing/ to obtain the image and annotation files and save them to .lst files.

  • From the initial MXNet download you should be able to find in the tools/ the file. Open a terminal and type:

$ cd /incubator-mxnet/tools/
$ python /project/path/DeepVisualAttributes /dataset/path/WIDER --quality=100 --pack-label=True

This will create the record files to wider_records/ to feed to the iterator.

Run the Code

  • Run Remember to provide as an input argumenet the data path.

If you use this code, please mention this repo and cite the paper:

author = {Sarafianos, Nikolaos and Xu, Xiang and Kakadiaris, Ioannis A.},
title = {Deep Imbalanced Attribute Classification using Visual Attention Aggregation},
booktitle = {ECCV},
year = {2018}


No description, website, or topics provided.



No releases published


No packages published