I saw your repo on using Barracuda in Unity with Yolo and felt inspired by it, I have got v2-tiny working with Unity using Barracuda with AR Foundation for AR object detection, however I am having difficulty understanding and reading Yolo v3 tiny output to create bounding boxes as it seems different.
I wondered if you would be interested on investigating using yolo v3 tiny with barracuda - I have made my repo open here https://github.com/ROBYER1/Unity_Detection2AR you are welcome to have a look and use it
The current plan is to share it with the Unity AR community as (hopefully) an official Unity Barracuda Object Detection example project once I have cracked v3-tiny
See a bit more about the progress I made here, currently I can pull the highest % detected classes from Yolo v3 tiny, but the logic for bounding boxes and confidence scores doesn't work the same for v3 as it did with v2 https://github.com/derenlei/Unity_Detection2AR/issues/9
Hey
I saw your repo on using Barracuda in Unity with Yolo and felt inspired by it, I have got v2-tiny working with Unity using Barracuda with AR Foundation for AR object detection, however I am having difficulty understanding and reading Yolo v3 tiny output to create bounding boxes as it seems different.
I wondered if you would be interested on investigating using yolo v3 tiny with barracuda - I have made my repo open here https://github.com/ROBYER1/Unity_Detection2AR you are welcome to have a look and use it
The current plan is to share it with the Unity AR community as (hopefully) an official Unity Barracuda Object Detection example project once I have cracked v3-tiny
See a bit more about the progress I made here, currently I can pull the highest % detected classes from Yolo v3 tiny, but the logic for bounding boxes and confidence scores doesn't work the same for v3 as it did with v2 https://github.com/derenlei/Unity_Detection2AR/issues/9
Kind Regards****