YOLO26N

[tcpipchip]

Hi

Is the git of sscma already updated ?

I am getting this error.

I trainned the yolo26 and got my cvimodel attached,
i updated the node-red node-red-contrib-sscma to 0.3.6

Is there something that i forgot ?

yolo26.zip

[LynnL4]

Sorry for the late reply. We have already started comparing the architecture of YOLOv26, and we will adapt to it as soon as possible.

[tcpipchip]

LynnL4

Don´t worry about DELAY.

Let me know when have progress! I am ready to test! I already created the yolo26 cvimodels to test with it.

If possible, add to the node-red the option to read a static image!

[LynnL4]

Hi, I’ve added support for YOLO26n. If you’d like to try it, you can compile it yourself for now. It’s expected to be officially released together with the reCamera Halo . Thanks.
Seeed-Studio/sscma-example-sg200x#39

[tcpipchip]

@LynnL4
We compiled...
Can you send us a yolo26 cvimodel to us ?, i want to test...they are not in this patch.
And will you give us new instructions to convert it to yolo26 cvimodel ?
Looks that will different of yolo11 instructions to convert
thank you!

[tcpipchip]

@LynnL4

Attached the

Custom .pt that i trainned using the yolo26n.pt as reference.

It´s the same for the MONKEY.

try to convert it to cvimodel and test on recamera!

best2626.zip

Thank you!!!!!!!!

[LynnL4]

Hi,
I used Claude to set up a fresh environment and convert the model, and everything seemed to work very smoothly. Maybe this will be helpful for you.
https://github.com/Seeed-Studio/ai-skills/tree/main/skills/onnx-to-cvimodel

yolo26n_cv181x_int8.zip

[tcpipchip]

Hi,

@LynnL4

Claude...now all on the Cloud ?

Did you test my cvimodel if opens there ? This is important for us.

Or got error ?

I tested your yolo26 ang got error here...

[tcpipchip]

I will do subscription to CLAUDE!

Nice AI.

and where do you execute the CLAUDE console ? Into the WSL2 + UBUNTU 22.0.4 ? or windows ?

and

witch is your sscma version ?

[root@milkv-duo]/usr/local/bin# sscma-node -v
0.2.2
build: Jan 29 2026 09:00:53

[tcpipchip]

hi, i am again :)

can you test my model if opens on your model recamera ?

yolov26_custom_cv181x_int8.zip

[tcpipchip]

i post an issue on

https://github.com/Seeed-Studio/ai-skills/tree/main/skills/onnx-to-cvimodel

can you verify that ?

[LynnL4]

hi, i am again :)

can you test my model if opens on your model recamera ?

yolov26_custom_cv181x_int8.zip

[tcpipchip]

Hi,
What is that 16 ?
i trainned 3 class...
Attached a new cvimodel

yolo11n-detect_cv181x_int8.zip

And a buggio image to you use as reference

When have time...test it...

[LynnL4]

As I don’t have your training dataset, the model converted to INT8 seems to have noticeable accuracy loss. The BF16 version looks much better in comparison — it might be worth trying that instead.

best2626_cv181x_bf16.zip