TIDL inference of YOLOV3 in ONNX input format

Hello Kumar,

Thanks for your reply.

Yes i have referred to the Validated YOLOV3 network and its prototxt file.

The ONNX file of YOLOV3 i wanted to infer is also having the same structure of layers (the.cfg file of model) and the same post processing activity as the TIDL validated network mentioned by you in the path.

But i have trained the network using PYTORCH framework, so does it works out if i train in mmdetection ?

PS:If i have scaled up the default values of the anchor boxes in the prototxt file then the BB's are of maximised sizes.

Can you please check the below parameters in your meta arch file 

message TIDLYoloParams {
required string input = 1;
repeated float anchor_width = 2;
repeated float anchor_height = 3;
}

Yes, the Yolo v3 model trained on mmdetection is validated.

Hello,

Can you mention the file name under which i can check these values.

As far as my model ill train it on mmdetection and i will try infering thanks for the info.