语义分割学习笔记3——运行语义分割程序

清疚 2022-10-08 11:16 612阅读 0赞

### 语义分割学习笔记3——运行语义分割程序 ###

*  一、linux安装依赖环境
 *   *  1、创建open-mmlab环境
     *  2、安装pytorch
     *  3、 安装 MMCV
     *  4、安装 MMSegmentation
 *  三、测试是否安装成功
 *  二、windows安装依赖环境

准备运行经典的**open-mmlab mmsegmentation**框架

**源码地址**：[https://github.com/open-mmlab/mmsegmentation][https_github.com_open-mmlab_mmsegmentation]  
安装说明：[https://github.com/open-mmlab/mmsegmentation/blob/master/README\_zh-CN.md][https_github.com_open-mmlab_mmsegmentation_blob_master_README_zh-CN.md]

# 一、linux安装依赖环境 #

cat /usr/local/cuda/version.txt

deepstream-app --version-all

> deepstream-app version 5.0.0  
> CUDA Runtime Version: 10.2

--------------------

## 1、创建open-mmlab环境 ##

这里我省略了。

## 2、安装pytorch ##

[语义分割学习笔记2——Jetson Xavier安装pytorch][2_Jetson Xavier_pytorch]

## 3、 安装 MMCV ##

pip install mmcv-full -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.6.0/index.html

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L21hb19odWlfZmVp_size_16_color_FFFFFF_t_70]

## 4、安装 MMSegmentation ##

pip install mmsegmentation

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L21hb19odWlfZmVp_size_16_color_FFFFFF_t_70 1]

# 三、测试是否安装成功 #

下载需要的模型

[https://github.com/open-mmlab/mmsegmentation/tree/master/configs][https_github.com_open-mmlab_mmsegmentation_tree_master_configs]  
下载下文源码中对应的cfg和checkpoint文件。

from mmseg.apis import inference_segmentor, init_segmentor
    import mmcv
    
    config_file = 'configs/pspnet/pspnet_r50-d8_512x1024_40k_cityscapes.py'
    checkpoint_file = 'checkpoints/pspnet_r50-d8_512x1024_40k_cityscapes_20200605_003338-2966598c.pth'
    
    # build the model from a config file and a checkpoint file
    model = init_segmentor(config_file, checkpoint_file, device='cuda:0')
    
    # test a single image and show the results
    img = 'test.jpg'  # or img = mmcv.imread(img), which will only load it once
    result = inference_segmentor(model, img)
    # visualize the results in a new window
    model.show_result(img, result, show=True)
    # or save the visualization results to image files
    # you can change the opacity of the painted segmentation map in (0, 1].
    model.show_result(img, result, out_file='result.jpg', opacity=0.5)
    
    # test a video and show the results
    video = mmcv.VideoReader('video.mp4')
    for frame in video:
       result = inference_segmentor(model, frame)
       model.show_result(frame, result, wait_time=1)

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L21hb19odWlfZmVp_size_16_color_FFFFFF_t_70 2]

或者：

python demo/image_demo.py demo/demo.jpg configs/pspnet/pspnet_r50-d8_512x1024_40k_cityscapes.py \
        checkpoints/pspnet_r50-d8_512x1024_40k_cityscapes_20200605_003338-2966598c.pth --device cuda:0 --palette cityscapes

# 二、windows安装依赖环境 #

最先是在windows下安装的，结果最后报：  
**ModuleNotFoundError: No module named 'mmcv.\_ext’**  
错误，一直安装不成功mmcv，官方现在似乎不提供windows下的mmcv的离线包，因此暂时放弃。（后面了解到windows安装版本只能是几个指定的版本）

[https_github.com_open-mmlab_mmsegmentation]: https://github.com/open-mmlab/mmsegmentation
[https_github.com_open-mmlab_mmsegmentation_blob_master_README_zh-CN.md]: https://github.com/open-mmlab/mmsegmentation/blob/master/README_zh-CN.md
[2_Jetson Xavier_pytorch]: https://blog.csdn.net/mao_hui_fei/article/details/117650584
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L21hb19odWlfZmVp_size_16_color_FFFFFF_t_70]: /images/20221005/1386a522753d45f09221af1b17a39b24.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L21hb19odWlfZmVp_size_16_color_FFFFFF_t_70 1]: /images/20221005/207c9d40f62f4f22a8506c4b7d0a1555.png
[https_github.com_open-mmlab_mmsegmentation_tree_master_configs]: https://github.com/open-mmlab/mmsegmentation/tree/master/configs
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L21hb19odWlfZmVp_size_16_color_FFFFFF_t_70 2]: /images/20221005/39d697d6e0cf48d3a665479a6a0d760a.png