行为识别 slowfast

原创

已于 2024-04-03 09:41:10 修改

· 284 阅读

4 ·

版权

文章标签：

#人工智能

于 2024-04-01 15:27:20 首次发布

概览

环境配置

slowfast源代码

GitHub - facebookresearch/SlowFast: PySlowFast: video understanding codebase from FAIR for reproducing state-of-the-art video models.

Installation

SlowFast/INSTALL.md at main · facebookresearch/SlowFast · GitHub

根据Requirements要求，编写对应的requirements.txt文件:

git+https://github.com/facebookresearch/fvcore
git+https://github.com/facebookresearch/fairscale
git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI
https://download.pytorch.org/whl/cu117/torchvision-0.15.2%2Bcu117-cp39-cp39-linux_x86_64.whl
https://download.pytorch.org/whl/cu117/torch-2.0.1%2Bcu117-cp39-cp39-linux_x86_64.whl
simplejson
av
iopath
psutil
opencv-python
tensorboard
moviepy
pytorchvideo
cython

ubuntu依赖：

1 GCC >= 4.9

使用命令：cat /proc/version 查看gcc版本

安装gcc的命令：

sudo apt install g++-4.9
sudo apt install gcc-4.9

拓展学习

cython

Cython 是什么？为什么会有 Cython？-优快云博客

pyav

Py之av：av库的简介、安装、使用方法之详细攻略_python av库-优快云博客

pip install git+

pip install git+-优快云博客

pip install -e

pip常用指令详解——（pip install -e .等）-优快云博客

run demo

执行demo的命令

python tools/run_net.py --cfg demo/AVA/SLOWFAST_32x2_R101_50_50.yaml

必要的代码：上一步Installation中下载的 detectron2_repo SlowFast-main

yaml配置改动

DEMO:
  ENABLE: True
  LABEL_FILE_PATH: /home/wengyang/codes/learn-slowfast/SlowFast-main/demo/AVA/class80_new.json # Add local label file path here.
  INPUT_VIDEO: /home/wengyang/codes/learn-slowfast/input/lancer.mp4
  OUTPUT_FILE: /home/wengyang/codes/learn-slowfast/output/lancer.mp4
  #  WEBCAM: 0
  DETECTRON2_CFG: "COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml"
  DETECTRON

最低0.47元/天解锁文章