This tutorial demonstrate an end to end video analytics example with OpenCV-DNN. The code includes few pipe stages.
This tutorial demonstrates how to run image classification application, while utilizing OpenCV DNN for inferencing.
Test contents are located in "samples/end2end_video_analytics/test_content" folder.
The stages to run the tutorial
$ cd /opt/intel/computer_vision_sdk/deployment_tools/model_downloader $ sudo python3 downloader.py –name ssd300
Build inference engine samples Open a new terminal. You can build from original inference_engine/samples folder with "sudo su" as well.
$ cp -r /opt/intel/computer_vision_sdk/deployment_tools/inference_engine/samples ~/Desktop $ cd ~/Desktop/samples $ source /opt/intel/computer_vision_sdk/bin/setupvars.sh
$ mkdir build; cd build; cmake .. -DCMAKE_BUILD_TYPE=Release; make
You can run image classification on an image or a video using a trained network with multiple outputs on Intel® Processors using the following command:
The application outputs out.h264 (h264 video elementary stream with bounding box/ class label/ accuracy rate on the objects, you can play this with "$ mplayer out.h264").
Upon the start-up of the demo application, it reads command line parameters and loads a network and an image or a video to the OpenCVDNN plugin. When the inference is done, the application will compose bounding boxes, class labels, and accuracy rates on the detected objects and encode it to h264 video elementary stream.