Home

Awesome

Introduction

Vedatad is a single stage temporal action detection toolbox based on PyTorch. Vedatad includes implementations of the following temporal action detection algorithms:

Features

License

This project is released under the Apache 2.0 license.

Installation

Requirements

We have tested the following versions of OS and softwares:

Install vedatad

a. Create a conda virtual environment and activate it.

conda create -n vedatad python=3.8.5 -y
conda activate vedatad

b. Install PyTorch and torchvision following the official instructions, e.g.,

conda install pytorch torchvision -c pytorch

c. Clone the vedatad repository.

git clone https://github.com/Media-Smart/vedatad.git
cd vedatad
vedatad_root=${PWD}

d. Install vedatad.

pip install -r requirements/build.txt
pip install -v -e .

Data preparation

Please follow specified algorithm in config/trainval to prepare data, for example, see detail in configs/trainval/daotad.

Train

a. Config

Modify some configuration accordingly in the config file like configs/trainval/daotad/daotad_i3d_r50_e700_thumos14_rgb.py

b. Train

tools/dist_trainval.sh configs/trainval/daotad/daotad_i3d_r50_e700_thumos14_rgb.py "0,1,2,3"

Test

a. Config

Modify some configuration accordingly in the config file like configs/trainval/daotad/daotad_i3d_r50_e700_thumos14_rgb.py

b. Test

CUDA_VISIBLE_DEVICES=0 python tools/test.py configs/trainval/daotad/daotad_i3d_r50_e700_thumos14_rgb.py weight_path

Contact

This repository is currently maintained by Hongxiang Cai (@hxcai), Yichao Xiong (@mileistone), Chenhao Wang (@C-H-Wong).

Credits

We got a lot of code from vedadet, thanks to Media-Smart.