https://github.com/zjcv/tsn
[ECCV 2016] Temporal Segment Networks: Towards Good Practices for Deep Action Recognition
https://github.com/zjcv/tsn
action-recognition pytorch temporal-segment-networks tsn video-recognition video-understanding
Last synced: 2 months ago
JSON representation
[ECCV 2016] Temporal Segment Networks: Towards Good Practices for Deep Action Recognition
- Host: GitHub
- URL: https://github.com/zjcv/tsn
- Owner: ZJCV
- License: apache-2.0
- Created: 2020-08-28T11:39:32.000Z (about 5 years ago)
- Default Branch: master
- Last Pushed: 2021-01-26T07:32:37.000Z (over 4 years ago)
- Last Synced: 2025-08-11T00:46:40.453Z (2 months ago)
- Topics: action-recognition, pytorch, temporal-segment-networks, tsn, video-recognition, video-understanding
- Language: Python
- Homepage:
- Size: 519 KB
- Stars: 7
- Watchers: 2
- Forks: 3
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
«TSN»复现了论文Temporal Segment Networks提出的视频分类模型
## 内容列表
- [内容列表](#内容列表)
- [背景](#背景)
- [安装](#安装)
- [使用](#使用)
- [主要维护人员](#主要维护人员)
- [致谢](#致谢)
- [仓库](#仓库)
- [论文](#论文)
- [参与贡献方式](#参与贡献方式)
- [许可证](#许可证)
## 背景
[Temporal Segment Networks: Towards Good Practices for Deep Action Recognition](https://arxiv.org/abs/1608.00859)是视频分类任务中的经典实现
## 安装
通过`requirements.txt`安装运行所需依赖
```
$ pip install -r requirements.txt
```
处理数据时需要额外安装[denseflow](https://github.com/open-mmlab/denseflow)，可以在[innerlee/setup](https://github.com/innerlee/setup)中找到安装脚本
## 使用
首先设置`GPU`和当前位置
```
$ export CUDA_VISIBLE_DEVICES=1
$ export PYTHONPATH=.
```
* 训练
```
# 训练UCF101
# 单GPU
$ python tools/train.py --config_file=configs/tsn_r50_ucf101_rgb_224x3_seg.yaml
# 多GPU
$ python tools/train.py \
--config_file=configs/tsn_r50_ucf101_rgb_224x3_seg.yaml \
--eval_step=1000 \
--save_step=1000 \
-g=
```
* 测试
```
# 单模态测试
$ python tools/test.py
$ python tools/test.py configs/tsn_r50_ucf101_rgb_224x3_seg.yaml outputs/tsn_r50_ucf101_rgb_224x3_seg.pth
# 多模态融合测试 - RGB + RGBDiff
$ python tools/fusion.py
$ python tools/fusion.py \
configs/tsn_r50_ucf101_rgb_224x3_seg.yaml \
outputs/tsn_r50_ucf101_rgb_224x3_seg.pth \
configs/tsn_r50_ucf101_rgbdiff_224x3_seg.yaml \
outputs/tsn_r50_ucf101_rgbdiff_224x3_seg.pth
```
## 主要维护人员
* zhujian - *Initial work* - [zjykzj](https://github.com/zjykzj)
## 致谢
### 仓库
* [lufficc/SSD](https://github.com/lufficc/SSD)
* [yjxiong/tsn-pytorch](https://github.com/yjxiong/tsn-pytorch)
* [open-mmlab/mmaction2](https://github.com/open-mmlab/mmaction2)
* [ facebookresearch/SlowFast](https://github.com/facebookresearch/SlowFast)
### 论文
```
@misc{wang2016temporal,
title={Temporal Segment Networks: Towards Good Practices for Deep Action Recognition},
author={Limin Wang and Yuanjun Xiong and Zhe Wang and Yu Qiao and Dahua Lin and Xiaoou Tang and Luc Van Gool},
year={2016},
eprint={1608.00859},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
## 参与贡献方式
欢迎任何人的参与！打开[issue](https://github.com/zjykzj/TSN/issues)或提交合并请求。
注意:
* `GIT`提交，请遵守[Conventional Commits](https://www.conventionalcommits.org/en/v1.0.0-beta.4/)规范
* 语义版本化，请遵守[Semantic Versioning 2.0.0](https://semver.org)规范
* `README`编写，请遵守[standard-readme](https://github.com/RichardLitt/standard-readme)规范
## 许可证
[Apache License 2.0](LICENSE) © 2020 zjykzj