安装最新版本
本笔记创建时,对应的最新版本是v1.2.0。
# 创建 conda 虚拟环境
conda create --name mmaction2 python=3.8 -y
conda activate mmaction2
# 安装 pytorch
pip install torch==1.10.1+cu102 torchvision==0.11.2+cu102 -f https://download.pytorch.org/whl/cu102/torch_stable.html
# 安装 mmaction2 依赖库
pip install -U openmim
mim install mmengine
mim install mmcv
mim install mmdet (optional)
mim install mmpose (optional)
# 安装 mmaction2
git clone https://github.com/open-mmlab/mmaction2.git
cd mmaction2
pip install -v -e .
# 验证安装是否成功
# 在 mmaction2 文件夹下执行
mim download mmaction2 --config tsn_imagenet-pretrained-r50_8xb32-1x1x8-100e_kinetics400-rgb --dest .
python demo/demo.py tsn_imagenet-pretrained-r50_8xb32-1x1x8-100e_kinetics400-rgb.py tsn_imagenet-pretrained-r50_8xb32-1x1x8-100e_kinetics400-rgb_20220906-2692d16c.pth demo/demo.mp4 tools/data/kinetics/label_map_k400.txt
安装v0.24.1版本
最新版本删除了 tools/deployment/pytorch2onnx.py
,无法将模型转换为 ONNX 格式,不便于部署。这个版本还保留有该文件。
可以使用mmdeploy实现pytorch模型到ONNX的转换。
# 创建 conda 虚拟环境
conda create -n mmaction2_v0.24.1 python=3.8 -y && conda activate mmaction2_v0.24.1
# 安装 pytorch
conda install numpy=1.23.1 # numpy 版本不能太高
conda install pytorch==1.10.0 torchvision==0.11.0 cudatoolkit=10.2 -c pytorch
# 安装 mmcv
pip install mmcv-full==1.4.8 -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.10.0/index.html
# 安装 mmaction2
git clone https://github.com/open-mmlab/mmaction2.git
cd mmaction2
git checkout v0.24.1 # 更换版本
pip install -r requirements/build.txt
pip install -v -e . # or "python setup.py develop"
# 安装其他库
pip install yapf==0.40.1
测试是否安装成功
import torch
from mmaction.apis import init_recognizer, inference_recognizer
config_file = 'configs/recognition/tsn/tsn_r50_video_inference_1x1x3_100e_kinetics400_rgb.py'
device = 'cuda:0' # or 'cpu'
device = torch.device(device)
model = init_recognizer(config_file, device=device)
# inference the demo video
inference_recognizer(model, 'demo/demo.mp4')
标签:demo,配置,环境,mmaction2,conda,install,pip,MMAction2,安装
From: https://blog.csdn.net/qq_42693593/article/details/142046204