Adaptive occlusion sensitivity analysis

This repository contains the implementation of the paper "Visually explaining 3D-CNN predictions for video classification with an adaptive occlusion sensitivity analysis".

This repository includes the work that is distributed in the Apache License 2.0.

Requirements

• docker > 20.
• docker-compose

Instalation

Download model parameters

1. Download the parameters of R3D fine-tuned on UCF-101 from here
2. Place the downloaded file (save_200.pth) into data/r3d_models/finetuning/ucf101/r3d50_K_fc/

Create docker container

1. $cd .devcontainer
2. $docker-compose up -d
3. $docker attach aosa

Example

Please refer to occlusion_sensitivity_analysis.ipynb.

If there is no enough GPU memory, please try to small "batchsize" in the example codes.

Models and dataset utils

We use the code from the following repository for 3D-CNN models and dataset utilities. To download datasets and other resources, please refer to this repository.

kenshohara/3D-ResNets-PyTorch: 3D ResNets for Action Recognition (CVPR 2018)

Citation

@article{uchiyama2022visually,
  title = {{Visually explaining 3D-CNN predictions for video classification with an adaptive occlusion sensitivity analysis}},
  author = {Uchiyama, Tomoki and Sogi, Naoya and Niinuma, Koichiro and Fukui, Kazuhiro},
  journal={arXiv preprint arXiv:2207.12859},
  year = {2022}
}