When running the OpenVINO sample using mobilenet-ssd model, there is no detection result

[rachel_r_Intel]

Hi All,

I want to verify the performance of  mobilenet-ssd model, here are steps:

/opt/intel/computer_vision_sdk/deployment_tools/model_downloader$ sudo python3 ./downloader.py --name mobilenet-ssd

/opt/intel/computer_vision_sdk/deployment_tools/model_downloader/object_detection/common/mobilenet-ssd/caffe$ sudo python3 /opt/intel/computer_vision_sdk/deployment_tools/model_optimizer/mo.py --input_model /opt/intel/computer_vision_sdk/deployment_tools/model_downloader/object_detection/common/mobilenet-ssd/caffe/mobilenet-ssd.caffemodel --output_dir /opt/intel/computer_vision_sdk/deployment_tools/model_downloader/object_detection/common/mobilenet-ssd/caffe/output/

/opt/intel/computer_vision_sdk/deployment_tools/inference_engine/samples/build/intel64/Release$ ./object_detection_demo_ssd_async -i cam -m /opt/intel/computer_vision_sdk/deployment_tools/model_downloader/object_detection/common/mobilenet-ssd/caffe/output/mobilenet-ssd.xml -d CPU

However, some common objects like person, chair,monitor etc are not detected. This is a bug, or because of the limitations of the type of mobilenet-ssd model detection?

Thanks!

[Cary_P_Intel1 · Accepted solution]

Hi, Rachel,

Since mobilenet-ssd requests to normalized the input data (to [-1, 1]), so you need to add extra parameters while converting the model as below (my environment is Windows, please change to your environment command)

python C:\Intel\computer_vision_sdk\deployment_tools\model_optimizer\mo.py --input_model object_detection\common\mobilenet-ssd\caffe\mobilenet-ssd.caffemodel --output_dir object_detection\common\mobilenet-ssd\caffe\IR --mean_values [127.5,127.5,127.5] --scale_values [127.5]

You can also refer to the document on why apply those values, please navigate to topic "When to Specify Mean and Scale Values"

https://docs.openvinotoolkit.org/2018_R5/_docs_MO_DG_prepare_model_convert_model_Converting_Model_General.html