目标检测库MMDetection的配置（MMDetection: Open MMLab Detection Toolbox and Benchmark）

 

1、论文下载地址：

MMDetection: Open MMLab Detection Toolbox and Benchmark. [paper] 

2、代码下载地址：

https://github.com/open-mmlab/mmdetection

3、创建虚拟环境并激活

conda create -n openmmlab python=3.7 -y
conda activate openmmlab

4、安装pytorch和torchvision

pip install torch===1.4.0 -f https://download.pytorch.org/whl/torch_stable.html
 
pip install torchvision===0.5.0 -f https://download.pytorch.org/whl/torch_stable.html

5、安装mmcv-full

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu100/torch1.4.0/index.html

我的CUDA版本是10.0，torch版本是1.4.0，大家针对性更改。安装编译需要很长一段时间，耐心等待。

6、安装apex

下载apex：

git clone https://github.com/NVIDIA/apex.git

 进入apex目录安装

cd apex

python setup.py install

7、安装MMDetection

进入第2步骤下载的工程路径运行：

pip install -r requirements/build.txt
python setup.py develop

可能遇到错误：

No local packages or working download links found for panopticapi@ git+https://github.com/cocodataset/panopticapi.git
error: Could not find suitable distribution for Requirement.parse('panopticapi@ git+https://github.com/cocodataset/panopticapi.git')

错误原因：

服务器上网络问题，连不上github，所以git库包panopticapi失败。

解决方法：

1）手动下载panopticapi包

https://github.com/cocodataset/panopticapi

2）放到工程路径解压

3）手动编译

 python panopticapi/setup.py develop

 4）打开runtime.txt，删除第三行的 @ git+https://github.com.cnpmjs.org/cocodataset/panopticapi.git

 5）重新运行python setup.py develop，编译成功。

8、下载预训练模型

下载地址：

https://github.com/open-mmlab/mmdetection/blob/master/docs/model_zoo.md

以faster R-CNN为例

1）

 2）

3）新建checkpoints路径，将下载的模型放入

9、新建demo.py文件，输入以下代码

from mmdet.apis import init_detector, inference_detector, show_result_pyplot
import cv2

config_file = 'configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py'
# download the checkpoint from model zoo and put it in `checkpoints/`
# url: https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth
checkpoint_file = 'checkpoints/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth'
device = 'cuda:0'
# init a detector
model = init_detector(config_file, checkpoint_file, device=device)
# inference the demo image

image='demo/demo.jpg'

result = inference_detector(model, image)

show_result_pyplot(model, image, result, score_thr=0.3)

# image = model.show_result(image, result, score_thr=0.3)
#
# cv2.imshow('demo', image)
# cv2.waitKey()

10、运行得到结果

python demo.py