Hide and Seek - Weakly Supervised Object Detector

Implementation of "Hide-and-Seek: Forcing a Network to be Meticulous for Weakly-supervised Object and Action Localization" https://arxiv.org/abs/1704.04232

Download data

Download data with the script below. The script download tiny-imagenet data and transform to pickle and tfrecord. The pickle include datasets and meta information of label dictionary. The tfrecord contains datasets; train dataset, valid dataset, test dataset.

python3 download_data.py

Requirements

• python3.4+
• pip3
• python packages - requirements.txt
• tensorflow 1.4+

How to train

$ jupyter lab

and open train_HaS.ipynb

or 

use snippet_train_model.py

How to test

$ jupyter lab

and open load_and_test.ipynb

Architecture

Hide Patches Example

Result Sample

• Green bounding box is prediction
• Red bounding box is ground truth