Intel® Distribution of OpenVINO™ Toolkit
Community support and discussions about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all things computer vision-related on Intel® platforms.

Faster RCNN troubleshoot on NCS2


I recently got my hand on both NCS and NCS2 and was able to convert and run SSD (custom and pretrained Models) from the tensorflow model zoo on both devices with an increase in speed (on one image inference) from 9.6 fps on the NCS to 14 fps on the NCS2

For the Faster-RCNN everything runs as expected on the NCS. However, moving to the NCS2 there is a problem while loading the model to the plugin, as it takes forever to load it (15 to 30 min depending on the demo executed). (log msg follows)

I also tried another version of Faster-RCNN VGG16 on Caffe. For this version the model was loaded normally however I observed a significant decrease in speed ( 22s per frame on NCS  vs 42s on the NCS2). Which doesnt really make sense as the NCS2 is supposed to be faster.

Did anyone already tried to run FASTER-RCNN on the NCS2. is it supported at all?


After loading the FasterRCNN Tensorflow Models on the NCS2 the performance and the speed is better observed for the same model on the NCS. 
Is there any difference in architecture between the two NCSs versions we should be aware of?

Any help would be really appreciated! 

root@CIATES-DEV11:~/inference_engine_samples_build/intel64/Release# sudo ./object_detection_demo_ssd_async -i /dev/video0 -m /home/ciates-dev11/Desktop/Ahmed/openvinoModels/FasterRCNN_InceptionV2.xml -d MYRIAD
    API version ............ 1.4
    Build .................. 19154
[ INFO ] Parsing input parameters
[ INFO ] Reading input

(object_detection_demo_ssd_async:14738): GStreamer-CRITICAL **: gst_element_get_state: assertion 'GST_IS_ELEMENT (element)' failed
[ INFO ] Loading plugin

    API version ............ 1.5
    Build .................. 19154
    Description ....... myriadPlugin
[ INFO ] Loading network files
[ INFO ] Batch size is forced to  1.
[ INFO ] Checking that the inputs are as the demo expects
[ INFO ] Checking that the outputs are as the demo expects
[ INFO ] Loading model to the plugin

0 Kudos
0 Replies