TemporalMaxer

TemporalMaxer copied to clipboard

TemporalMaxer: Maximize Temporal Context with only Max Pooling for Temporal Action Localization

This is the official implementation of the paper TemporalMaxer.

We release the training and testing code for THUMOS, EPIC-Kitchen 100 (verb, noun), and MultiTHUMOS datasets.

Operating Systems and Specs

• Ubuntu 18.04.5 LTS
• NVIDIA RTX A6000
• NVIDIA-SMI 520.61.05 - Driver Version: 520.61.05 - CUDA Version: 11.8

Table of Contents

1. Introduction
2. Installation
3. Usage
4. Citation

Introduction

Recent studies have emphasized the importance of applying long-term temporal context modeling (TCM) blocks to the extracted video clip features such as employing complex self-attention mechanisms. In this paper, we present the simplest method ever to address this task and argue that the extracted video clip features are already informative to achieve outstanding performance without sophisticated architectures. To this end, we introduce TemporalMaxer, which minimizes long-term temporal context modeling while maximizing information from the extracted video clip features with a basic, parameter-free, and local region operating max-pooling block. Picking out only the most critical information for adjacent and local clip embeddings, this block results in a more efficient TAL model. We demonstrate that TemporalMaxer outperforms other state-of-the-art methods that utilize long-term TCM such as self-attention on various TAL datasets while requiring significantly fewer parameters and computational resources.

Installation

a. Install packages

conda create -n TemporalMaxer python=3.9
conda activate TemporalMaxer
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch -y
python -m pip install -r requirements.txt
pip install -e ./

b. Build NMS

Part of NMS is implemented in C++. The code can be compiled by

cd ./libs/utils; python setup.py install; cd ../..

The code should be recompiled every time you update PyTorch.

Usage

Reproduce Our Results on EPIC-Kitchens 100

Download Features and Annotations

• Download epic_kitchens.tar.gz from this link, md5sum 6cbf312eb5025c0abbbf5d0eaa61e556.
• Make data folder in the current code directory.
• The data folder structure should look like the following:

# This folder
├── configs
│   ├── temporalmaxer_epic_slowfast_noun.yaml
│   └── temporalmaxer_epic_slowfast_verb.yaml
│   └── ........
├── data 
│   ├── epic_kitchens
│       ├── annotations
│       └── features
├── eval.py
├── figures
├── libs
    ........

• Train and test EPIC-Kitchens 100 VERB

# training
./scripts/epic_verb/train.sh
# testing
./scripts/epic_verb/test.sh

• Train and test EPIC-Kitchens 100 NOUN

# training
./scripts/epic_noun/train.sh
# testing
./scripts/epic_noun/test.sh

• The results should be:

Reproduce Our Results on THUMOS

Download Features and Annotations

• Download thumos.tar.gz from this link, md5sum 1f71c37dba55d549e4b02841d0dcf603.
• Make data folder in the current code directory.
• The data folder structure should look like the following:

# This folder
├── configs
│   └── ........
│   └── temporalmaxer_thumos_i3d.yaml
├── data
│   ├── thumos
│       ├── annotations
│       └── i3d_features
├── eval.py
├── figures
├── libs
    ........

• Train and test THUMOS

# training
./scripts/thumos/train.sh
# testing
./scripts/thumos/test.sh

• The results should be:

Reproduce Our Results on MultiTHUMOS

Download Features and Annotations

• Download multithumos.tar.gz from this link, md5sum 5b2477678abd612440b6099e349442ad.
• Make data folder in the current code directory.
• The data folder structure should look like the following:

# This folder
├── configs
│   └── ........
│   └── temporalmaxer_multithumos_i3d.yaml
│   └── ........
├── data
│   ├── thumos
│   │   ├── annotations
│   │   └── i3d_features
│   ├── multithumos
│       ├── ........
│       └── multithumos.json
│       └── ........
├── eval.py
├── figures
├── libs
    ........

• Train and test MultiTHUMOS

# training
./scripts/multithumos/train.sh
# testing
./scripts/multithumos/test.sh

• The results should be:

Citation

Please cite the paper in your publications if it helps your research:

@article{tang2023temporalmaxer,
  title={TemporalMaxer: Maximize Temporal Context with only Max Pooling for Temporal Action Localization},
  author={Tang, Tuan N and Kim, Kwonyoung and Sohn, Kwanghoon},
  journal={arXiv preprint arXiv:2303.09055},
  year={2023}
}