How to deploy my model to VTA?


#1

Hello, everyone!
1.If I want to put my model that I’ve trained into VTA, how should I do?

2.If I want to put my quantized model that I’ve trained by tensorflow into VTA, how should I do?

3.Which method above are better?

4.Can anyone give me some detailed example?

Thank you very much!


#2

I think YoloV2 models may not be supported for import now on Tensorflow frontend.
Hence left with Darknet only for now.

About deployment, you may refer to from_darknet.py for compilation followed by VTA tutorial to customise it for your target.


#3

We are actively working on YOLOv3 support on VTA; the first version will likely be INT8. What precision does your model use?

CC @jroesch @thierry


#4

I might use Darknet to train it now as @srkreddy1238 say above, but I am not sure the precision about Darknet’s model.


#6

cc @siju-samuel @PariksheetPinjari909 can you share the details on Yolo with darknet ?


#7

https://pjreddie.com/darknet/yolov2/
There are files in that website. I use these two files.%E5%9C%96%E7%89%87


#8

I might use Darknet to train it now as @srkreddy1238 say above, but I am not sure the precision about Darknet’s model.