# Inference Engine Docker
## 0x01 Dependencies
### 1.1 Deepstream
> 请参考 [Deepstream User Guide](https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_Quickstart.html).
**以下实现针对 Tesla V100s 服务器，确保宿主机满足 NVIDIA driver 470.63.01 或更高，需要宿主机 root 权限.**
### 1.2 Docker
**a. 从 Dockerfile 创建镜像**
```
# 在当前路径下执行
docker build -t your-image-name:image-tag .
```
**b. 启动并进入容器**
```
docker run -itd --gpus all --name your-container-name -v /your/path-to/inference-engine/:/inference-engine/ --net host your-image-name:your-image-tag
docker exec -it your-container-name bash
```
**c. 一键拉取并完成 python binding**
> 参考官方样例 [Deepstream Python apps](https://github.com/NVIDIA-AI-IOT/deepstream_python_apps.git)
```
cd /deepstream-inference-engine
bash ./test-sample.sh
```
**完成以上步骤后，可以通过以下 Quickstart 运行 nvidia 官方测试样例进行测试，或者直接参考主目录 Readme，从 1.3 YOLO v5 继续执行跑通 inference engine**
## 0x02 Quickstart
### 2.1 启动样例 rtsp 服务
```
cd /opt/nvidia/deepstream/deepstream-6.0/samples/deepstream_python_apps/apps/deepstream-rtsp-in-rtsp-out
python3 deepstream_test1_rtsp_in_rtsp_out.py -i rtsp://xxxx
```
### 2.2 本地查看
* 使用 VLC 或 ffmpeg 通过网络串流展示.
## 0x03 ToubleShooting
1. `CAfile: /etc/ssl/certs/ca-certificates.crt CRLfile: none`
* 解决方法：`export GIT_SSL_NO_VERIFY=1`
2. `No package 'gstreamer-1.0' found`
* 解决方法：`apt-get install libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev`
3. `ValueError: Namespace GstRtspServer not available`
* 解决方法：`apt-get install gir1.2-gst-rtsp-server-1.0`
4. `ERROR: ...... failed because file path: ...... open error`
* 解决方法：确保报错路径中包含所需文件（首次尝试 Quickstart 建议拷贝 `test-sample.sh` 至 `/opt/nvidia/deepstream/deepstream-6.0/samples/` 并在该路径下执行后续操作）
基于 Deepstream 的深度学习多路视频推理平台.zip
版权申诉
157 浏览量
2024-02-19
15:44:14
上传
评论
收藏 66KB ZIP 举报
基于 Deepstream 的深度学习多路视频推理平台.zip （63个子文件） 
deepstream-inference-engine-master 
app.py 2KB Docker Dockerfile 573B test-sample.sh 421B README.md 2KB configs official-yolov5n 
deepstream_app_config.txt 870B yolov5n.pt 132B labels.txt 624B config_infer_primary_yoloV5.txt 817B grpc inference_result.proto 605B phone-call-detect deepstream_app_config.txt 832B phone-call-detect.pt 133B labels.txt 16B config_infer_primary_yoloV5.txt 837B dataset 
sample_qHD.mp4 133B test-phone-call.mp4 133B sample_720p.mp4 133B core probe probe_base.py 3KB phone_call_detect_probe.py 2KB jam_detect_probe.py 2KB engine.py 16KB common __init__.py 0B stupid_global_var.py 246B draw_bounding_boxes.py 967B grpc_client.py 1KB thirdparty nvdsinfer_custom_impl_Yolo nvdsinfer_yolo_engine.cpp 4KB layers route_layer.cpp 2KB upsample_layer.cpp 773B dropout_layer.cpp 287B maxpool_layer.h 372B activation_layer.h 446B shortcut_layer.h 446B shortcut_layer.cpp 1KB upsample_layer.h 393B channels_layer.cpp 781B implicit_layer.h 417B activation_layer.cpp 4KB dropout_layer.h 295B implicit_layer.cpp 786B convolutional_layer.cpp 7KB channels_layer.h 371B maxpool_layer.cpp 921B route_layer.h 373B convolutional_layer.h 574B yoloForward.cu 4KB yoloForward_r.cu 3KB yolo.h 3KB yolo.cpp 24KB Makefile 3KB yoloPlugins.cpp 8KB calibrator.h 2KB utils.cpp 5KB nvdsparsebbox_Yolo.cpp 14KB yoloPlugins.h 6KB yoloForward_v2.cu 3KB calibrator.cpp 5KB utils.h 2KB yoloForward_nc.cu 3KB README.md 212B scripts gen_wts_yoloV5.py 15KB flask_remove_source.sh 69B flask_add_source.sh 284B rtsp_server.py 2KB README.md 780B资源评论
