deepstream运行TAO模型

nvidia TAO有很多预训练模型，如：
##### Detection Network

- Faster-RCNN / YoloV3 / SSD / DSSD / RetinaNet / YoloV4  (https://ngc.nvidia.com/catalog/models/nvidia:tao:Faster-RCNN)
- DetectNet_v2 (https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/pretrained_detectnet_v2)

##### Classification Network

- multi_task (https://ngc.nvidia.com/catalog/models/nvidia:tao:pretrained_image_classification)

##### Other Networks

- DashCamNet (https://ngc.nvidia.com/catalog/models/nvidia:tao:dashcamnet)
- VehicleMakeNet (https://ngc.nvidia.com/catalog/models/nvidia:tao:vehiclemakenet)
- VehicleTypeNet (https://ngc.nvidia.com/catalog/models/nvidia:tao:vehicletypenet)
- TrafficeCamNet (https://ngc.nvidia.com/catalog/models/nvidia:tao:trafficcamnet)
- PeopleNet (https://ngc.nvidia.com/catalog/models/nvidia:tao:peoplenet)
- FaceDetectIR (https://ngc.nvidia.com/catalog/models/nvidia:tao:facedetectir)
- PeopleSegNet (https://ngc.nvidia.com/catalog/models/nvidia:tao:peoplesegnet)
- PeopleSemSegnet (https://ngc.nvidia.com/catalog/models/nvidia:tao:peoplesemsegnet)

deepstream_reference_apps
可以把这些模型放到deepstream中测试，步骤如下：
1 获取配置文件
$ git clone https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps.git
$ cd deepstream_reference_apps/deepstream_app_tao_configs/
$ sudo cp -a * /opt/nvidia/deepstream/deepstream/samples/configs/tao_pretrained_models/

2 下载模型
$ sudo apt install -y wget zip
$ cd /opt/nvidia/deepstream/deepstream/samples/configs/tao_pretrained_models/
$ sudo ./download_models.sh
 
3 运行Yolo
./deepstream-app  -c ../../../../samples/configs/tao_pretrained_models/deepstream_app_source1_detection_models.txt

deepstream-test3运行yolov4-tiny

1 从https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/blob/master/deepstream_app_tao_configs/download_models.sh 下载Model， 也可直接执行 wget https://nvidia.box.com/shared/static/em2dh1h4isjhfu7qf0hh6ggzbusdg29 -O models.zip

2 下载配置文件https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/blob/master/deepstream_app_tao_configs/config_infer_primary_yolov4-tiny_jetson.txt

和标签文件https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/blob/master/deepstream_app_tao_configs/yolov4_tiny_labels.txt

将两个文件放到deepstream-test3目录，并修改配置文件的模型路径。

3 执行例子：./deepstream-test3-app rtsp://127.0.0.1:8554/test

deepstream_tao_apps

./apps/tao_segmentation/ds-tao-segmentation -c configs/peopleSemSegNet_tao/shuffle/pgie_peopleSemSegShuffleUnet_tao_config.txt  -i file:///home/sample_720_2s.mp4