Initial commit for AXERA AI accelerators #20685
Conversation
✅ Deploy Preview for frigate-docs ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
|
Hi, support for this hardware has already been added in #19680 |
Modify AXERA section and related content from hardware documentation.
|
Hi, this detector uses the AXERA AX650N/AX8850 NPU, which is different hardware from the Synaptics SL1680 NPU. Please review it again. |
|
We can't accept PRs that include references to AGPL licensed models. See here |
|
Thank you. We will remove the AGPL licensed YOLO and submit a new, compliant model. |
|
In the previous commit, we updated the default object detection model to YOLOv9. |
|
Is there an example of converting a YOLOv9 ONNX model to AXModels? This might be particularly useful for Frigate Plus users. |
The model is now dynamically downloaded to the cache directory. Post-processing is now done using Frigate's built-in `post_process_yolo`. Configuration in the relevant documentation has been updated.
ivanshi1108
force-pushed
the
AXERA-axcl
branch
from
95a3fe2 to
1dee548
Compare
NickM-27
left a comment
NickM-27
left a comment
There was a problem hiding this comment.
A few notes and the python checks are failing
Co-authored-by: Nicolas Mowen <[email protected]>
5 participants
add a detector which is testing with yolov9 axmodel.
Proposed change
Add a detector using the AXERA AI accelerator.
As discussed in the linked thread
we want to add a new type of detector, namely the AXERA NPU. The current default detection model is YOLOv5
This initial commit serves as the beginning of the upcoming work.
Type of change
Additional information
Checklist
enlocale.ruff format frigate)