MIST: Multiple Instance Spatial Transformer Network

Baptiste Angles, Yuhe Jin, Simon Kornblith, Andrea Tagliasacchi, Kwang Moo Yi

This repository contains training and inference code for MIST: Multiple Instance Spatial Transformer Network.

Installation

This code is implemented based on PyTorch. A conda environment is provided with all the dependencies:

conda env create -f system/conda_mist.yaml

Pretrained models and datasets

Two pretrained models are provided for MNIST dataset and trimmed Pascal+COCO dataset respectively. Models download path:

mkdir pretrained_models
wget https://www.cs.ubc.ca/research/kmyi_data/files/2021/mist/mnist_best_models -P ./pretrained_models/
wget https://www.cs.ubc.ca/research/kmyi_data/files/2021/mist/pascal_coco_best_models -P ./pretrained_models/

Dataset download path:

mkdir dataset
wget https://www.cs.ubc.ca/research/kmyi_data/files/2021/mist/mnist_hard.zip -P ./dataset/
wget https://www.cs.ubc.ca/research/kmyi_data/files/2021/mist/VOC_pascal_coco_v2.zip -P ./dataset/
unzip ./dataset/mnist_hard.zip -d ./dataset/
unzip ./dataset/VOC_pascal_coco_v2.zip -d ./dataset/

Inference

Following commands will run pretrained model on test set. Visualization can be found in './test_results'

python mist_test.py --path_json='json/pascal.json'
python mist_test.py --path_json='json/mnist.json'

Citation

@inproceedings{angles2021mist,
  title={MIST: Multiple Instance Spatial Transformer Networks},
  author={Baptiste Angles*, Yuhe Jin*, Simon Kornblith, Andrea Tagliasacchi, Kwang Moo Yi},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  year={2021}
}

ubc-vision / mist