YOLOV9-T Compatibility with AMB82-MINI

Technical Discussion
1

Hello,

I’ve been working with the AMB82-MINI platform for object detection tasks and would like to confirm whether it supports the YOLOv9-t model, trained using the following repository: GitHub - WongKinYiu/yolov9: Implementation of paper - YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information .

Looking forward to your guidance :).

BR,
Hari

2

Hi @harir ,

May I know if you are using standard SDK or Arduino SDK?

yolov9 tiny models are supported on Standard SDK, ameba-rtos-pro2/project/realtek_amebapro2_v0_example/src/mmfv2_video_example/mmf2_video_example_vipnn_rtsp_init.c at 515dfc53d5ad7bd440f7a40b9978fdff7db60580 · Ameba-AIoT/ameba-rtos-pro2 · GitHub

1 Like
3

Hello @pammyleong,

Thanks for your reply.

I am using the Arduino SDK.

4

Hi @harir

Currently YOLOv9 tiny is not yet supported on Arduino SDK. (Only YOLOv3, v4 and v7 tiny supported).

May I know if your project requires the use of YOLOv9tiny?

Thank you.

5

Hello @pammyleong

My project does not require YOLOv9Tiny, particularly. So far, I have been working with YOLOV7 (repo from Wongkinyu) to train my model and deploy it in AMB82-MINI (using the online converter tool). But, when trying to convert it offline (using NN Deployment — AmebaPro2's Documentation v0.1 documentation ) I am facing an issue after making the changes in the respective import.sh file.

I have created a new topic YOLOv7 .pt to .nb Local conversion and deployement for AMB82-MINI – No Bounding Box or Score Output regardin this.

BR, Hari