https://github.com/milkcat0904/temporal-segment-network-pytorch
tsn model for action recognition on pytorch
https://github.com/milkcat0904/temporal-segment-network-pytorch
action-recognition python-3-6 pytorch temporal-segment-networks
Last synced: 17 days ago
JSON representation
tsn model for action recognition on pytorch
- Host: GitHub
- URL: https://github.com/milkcat0904/temporal-segment-network-pytorch
- Owner: milkcat0904
- Created: 2017-12-29T02:04:35.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2017-12-29T07:10:45.000Z (over 7 years ago)
- Last Synced: 2025-03-23T19:23:10.807Z (about 1 month ago)
- Topics: action-recognition, python-3-6, pytorch, temporal-segment-networks
- Language: Python
- Size: 45.2 MB
- Stars: 17
- Watchers: 0
- Forks: 4
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Temporal Segment Networks
tsn model for action recognition on pytorch
1.编译环境 Build
=
build opencv 2.4.13 和 dense_flow 环境
$ bash build_all.sh
终端输出"All tools built. Happy experimenting!" 表示build成功
2.下载数据集 Download dataset
=
实验数据集采用UCF-101，HMDB51
* [ucf101][UCF101]
* [hmdb51][HMDB51]
3.提取光流 Extract optical flow
=
$ bash scripts/extract_optical_flow.sh DATASET_PATH OUT_PATH NUMBER_OF_WORKER
DATASET_PATH : 数据集的地址
OUT_PATH ：生成光流图的地址
NUMBER_OF_WORKER ：工作的显卡数量，一般设置为2
一个频频的光流图和frame和放在同一个文件夹下
4.提取warped光流图 Etract warped flow
=
* 将`tools/build_of.py`的第89行`--flow_type`的默认值改为`warped_tv11`：
```python
parser.add_argument("--flow_type", type=str, default='warped_tvl1', choices=['tvl1', 'warp_tvl1'])
```
* $ bash scripts/extract_optical_flow.sh DATASET_PATH OUT_PATH NUMBER_OF_WORKER
5.生成标签 Label
=
* ucf101数据集标签：
$ bash scripts/build_file_list.sh ucf101 FRAME_PATH
* hmdb51数据集标签：
$ bash scripts/build_file_list.sh hmdb51 FRAME_PATH
FRAME_PATH：光流图（frame）的位置
6.训练(Inception-BN) Training
=
* ucf101
* rgb模型
* flow模型
* rgb-diff模型
* warped flow模型
* hmdb51
* rgb模型
$ python main.py hmdb51 RGB
--arch BNInception --num_segments 3
--gd 20 --lr 0.001 --lr_steps 30 60 --epochs 80
-b 128 -j 8 --dropout 0.8
--snapshot_pref ucf101_bninception_
--gpus 0 1
* flow模型
* rgb-diff模型
* warped flow模型
[ucf101]:http://crcv.ucf.edu/data/UCF101.php
[hmdb51]:http://serre-lab.clps.brown.edu/resource/hmdb-a-large-human-motion-database/