Skip to content

Latest commit

 

History

History
64 lines (51 loc) · 1.96 KB

README.md

File metadata and controls

64 lines (51 loc) · 1.96 KB

CTRNet

The code of "Content Temporal Relation Network for Temporal Action Proposal Generation"

Contents


Usage Guide

Prerequisites

The training and testing in CTRNet is reimplemented in PyTorch for the ease of use.

  • [PyTorch 1.8.0][pytorch]
  • [Fastai 1.0.61][fastai]

Code and Data Preparation

Get the code

Clone this repo with git, please remember to use --recursive

git clone --recursive https://github.com/YanZhang-bit/CTRNet

Download Features

THUMOS14 feature can be download here.

ActivityNet-1.3 feature can be download here. And people need rescaled the feature length of all videos to same length 100.

Training CTRNet

  • To train our CTRNet on the THUMOS14 dataset
cd Thumos
python fastai_train.py
  • To train our CTRNet on the ActivityNet-1.3 dataset
cd ActivityNet
python fastai_train_content.py

Testing CTRNet

  • To test our CTRNet on the THUMOS14 dataset

    We provide the final model for THUMOS14 dataset Baidu Cloud (password: eyy6)

cd Thumos
python fastai_inference.py --checkpoint best_model/final_model.pth
python postprocess.py
python eval.py
  • To test our CTRNet on the ActivityNet-1.3 dataset

    We provide the final model for ActivityNet-1.3 dataset Baidu Cloud (password: 3qtc)

cd ActivityNet
python inference_content.py --mode inference --checkpoint best_model/final_model.pth