Deepstream测试自定义样例模型

1.基础依赖部署

1.1.安装tensorflow

• sudo pip install tensorflow --ignore-installed

1.2.安装tensorrt

• 下载tensorrt的开发包
• sudo pip2 install ./python/tensorrt-5.0.2.6-py2.py3-none-any.whl
• sudo pip2 install ./uff/uff-0.5.5-py2.py3-none-any.whl
• sudo pip2 install ./graphsurgeon/graphsurgeon-0.3.2-py2.py3-none-any.whl
• export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:<PATH_TO_TENSORRT>/lib

2.测试objectDetector_SSD

2.1.下载ssd模型包

• wget http://download.tensorflow.org/models/object_detection/ssd_inception_v2_coco_2017_11_17.tar.gz
• tar -zxvf ssd_inception_v2_coco_2017_11_17.tar.gz

2.2.转uff模型

• cp <PATH_TO_TENSORRT>/samples/sampleUffSSD/config.py .
• python2 /usr/local/lib/python2.7/dist-packages/uff/bin/convert_to_uff.py --input-file frozen_inference_graph.pb -O NMS -p config.py
• cp frozen_inference_graph.uff <PATH_TO_DEEPSTREAM>/sources/objectDetector_SSD/sample_ssd_relu6.uff
• cp <PATH_TO_TENSORRT>/data/ssd/ssd_coco_labels.txt <PATH_TO_DEEPSTREAM>/sources/objectDetector_SSD/ssd_coco_labels.txt

2.3.测试指令

gst-launch-1.0 filesrc location=../../samples/streams/sample_720p.mp4 ! decodebin ! nvinfer config-file-path= config_infer_primary_ssd.txt ! nvvidconv ! nvosd ! nveglglessink

3.测试FasterRCNN

3.1.下载faster_rcnn模型包

• wget https://dl.dropboxusercontent.com/s/o6ii098bu51d139/faster_rcnn_models.tgz
• tar -zxvf faster_rcnn_models.tgz

3.2.部署模型

• cp <PATH_TO_TENSORRT>/data/faster-rcnn/faster_rcnn_test_iplugin.prototxt .
• cp faster_rcnn_models/VGG16_faster_rcnn_final.caffemodel .

3.3.测试指令

gst-launch-1.0 filesrc location=../../samples/streams/sample_720p.mp4 ! decodebin ! nvinfer config-file-path= config_infer_primary_fasterRCNN.txt ! nvvidconv ! nvosd ! nveglglessink