Temporal Context Network for Activity Localization in Videos

This is the implementation for ICCV 17 paper "Temporal Context Network for Activity Localization in Videos".
If you use the code, pretrained models, proposals, please cite:

@InProceedings{Dai_2017_ICCV,
author = {Dai, Xiyang and Singh, Bharat and Zhang, Guyue and Davis, Larry S. and Qiu Chen, Yan},
title = {Temporal Context Network for Activity Localization in Videos},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {Oct},
year = {2017}
}

Pre-trained Proposals

We provide the pre-trained proposal for both ActivityNet and THUMOS to assist future temporal detection works.

Dataset	Link
ActivityNet	Download
THUMOS	Download

Run the code

Prerequisite: A caffe with python support

Set PYTHONPATH to pycaffe path
Set ACTNET_HOME to folder with features

run "all_in_one.sh" to train and test

Pre-trained Features

We fine-tune TSN on dataset and extract score features.

Dataset	Link
ActivityNet	Download1 Download2 Download3
THUMOS	Download

For global features such as mbh and imagenet_shuffle, you can download from the official website.

Pre-trained Models

Here are the pre-trained models:

Dataset	Link
ActivityNet	Download
THUMOS	Download