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.
5919 Discussions

Error running trained, converted darknet yolov3-tiny model on NCS2

idata
Community Manager
512 Views

We successfully trained a custom yolov3-tiny model on darknet (tested in darknet and confirmed to detect our single class). We successfully converted the weights to a .pb and then .bin and .xml files. The issue comes when trying to run this model on our modified version of the the openvino_tiny-yolov3_multiStick_test.py. We have all the config stuff set up to yolov3-tiny (classes = 1, num = 6, LABELS = "our_object"). When we run it, we get an index out of range error in predict_async

 

cnt,dev = heapq.heappop(self.heap_request)

 

HOWEVER if we change num=3 it runs but the output boxes labeled detections is garbage data. Any ideas how to fix this?
0 Kudos
1 Reply
idata
Community Manager
165 Views

Hi @davidtrost

 

Which application are you using? Can you share your code so I can also see the modifications you made please? I will try to reproduce this and get back to you with results.

 

Best Regards,

 

Sahira
Reply