Objectdetection

Development environment

UP2 board, myraid chip(AI core), Ubuntu 16.04

1.  Excute ‘build.sh’ in the archive file to compile the code.
   

2.  Download the video file into network/ir_fp32 or ir_fp16 folders.
   
    cd network/ir_fp32 # or ir_fp16
   
    wget https://raw.githubusercontent.com/nealvis/media/master/traffic_vid/bus_station_6094_960x540.mp4
   

3.  Download the weights file and convert the bin files.
   
    cd network
   
    a.      download  the SSD_GoogleNetV2 archive file from https://software.intel.com/file/609199/download =>SSD_GoogleNetV2_caffe.tgz
   
    b.     Uncompress the  archive files and rename SSD_GoogleNetV2.prototext from SSD_GoogleNetV2_Deploy.prototxt
   
    c.      Convert the caffemodel as OpenVINO format.
   
    /opt//intel/computer_vision_sdk_2018.1.249/deployment_tools/model_optimizer/mo.py --input_model SSD_GoogleNetV2.caffemodel --output_dir ir_fp16 --data_type FP16
   
   /opt//intel/computer_vision_sdk_2018.1.249/deployment_tools/model_optimizer/mo.py --input_model SSD_GoogleNetV2.caffemodel --output_dir ir_fp32 --data_type FP32
   

4.  Link the library file into ir_fp32 and ir_fp16 folders and the binary file in build folder.
   
    cd network/ir_fp32
   
    ln –s /opt/intel/computer_vision_sdk/deployment_tools/inference_engine/lib/ubuntu_16.04/ lib
   
    ln –s ../../build/intel64/Release/object_detection_demo_ssd_async   object_detection_demo_ssd_async
   

5.  Execute the demo
   
    a.      CPU mode
   
            cd network/ir_fp32
   
            ./object_detection_demo_ssd_async -i bus_station_6094_960x540.mp4 -m SSD_GoogleNetV2.xml 
   
    b.     MYRIAD mode 
   
             cd network/ir_fp16
   
             sudo ./object_detection_demo_ssd_async -i bus_station_6094_960x540.mp4 -m SSD_GoogleNetV2.xml -d MYRIAD