Features
Install mmdetection (2D object detection library) in an Ascend NPU container
Prerequisites
- mmcv is installed
- Ascend CANN and torch_npu are installed in the container
- The working directory is mounted
Usage
Step 1: Confirm the prerequisites
Ensure mmcv is correctly installed:
python -c "import mmcv; print(mmcv.__version__)"
Step 2: Clone the mmdetection repository
Choose according to the network conditions:
Direct server connection:
git clone -b v2.24.0 https://github.com/open-mmlab/mmdetection.git
Server network timeout: configure a proxy or clone locally and transfer via scp
# 方案1:配置代理
export http_proxy=http://proxy:port
export https_proxy=http://proxy:port
git clone -b v2.24.0 https://github.com/open-mmlab/mmdetection.git
# 方案2:本地克隆后传输(仍使用GitHub链接)
# 本地执行
git clone -b v2.24.0 https://github.com/open-mmlab/mmdetection.git
scp -r mmdetection user@server:/target/path/
Important: Always use the official GitHub link, do not use mirror sites.
Step 3: Apply patches (if any)
cd mmdetection
cp /path/to/DrivingSDK/model_examples/BEVFormer/mmdet_config.patch .
git apply --reject mmdet_config.patch
Step 4: Install
source /opt/conda/bin/activate <conda-env-name>
pip install -e .
Step 5: Verify the installation
python -c "import mmdet; print(mmdet.__version__)"
NPU verification:
python -c "import site; print([p for p in site.getsitepackages() if 'cuda' in p.lower()])"
# 应输出空列表
FAQ
- Missing dependencies: ensure mmcv and mmengine are installed
- Patch failure: skip non-critical patches or handle them manually
Decision guidance
Ask the user:
- Is mmcv installed?
- Can the server network access GitHub?
- Do specific patches need to be applied?