Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

OpenCV is a generic image processing library (with DNN support), OpenVINO is a kind of DNN compiler for Intel VPUs.

"OpenCV can use OpenVINO backend" means that OpenVINO has some OpenCV compatibility, so it can take OpenCV Mats and process them in the VPU. It doesn't mean that it can take OpenCV deep neural networks.

The problem with OpenVINO is that it needs to "translate" the Caffe/TF networks to a Myriad-compatible format. As OpenVINO is not complete yet, it can't understand every kind of DNN layer or parameter; so it will give a compilation error for some complex networks (as YoloV3).

The solution would be to use a similar network that OpenVINO understands: The ...-coco networks can be used as YOLO alternatives.