The issue about "yolov8n_nms_config.json" layer name

General
1

The issue is when open this file, l see the reference to “yolov8n/conv41” and other layers but they don’t correspond anymore to my custom model’s output layer’s name read from Netron.
The node name of my model are similar to the name in yolov8n.yaml file
“”- /model.22/cv2.0/cv2.0.2/Conv
- /model.22/cv3.0/cv3.0.2/Conv
- /model.22/cv2.1/cv2.1.2/Conv
- /model.22/cv3.1/cv3.1.2/Conv
- /model.22/cv2.2/cv2.2.2/Conv
- /model.22/cv3.2/cv3.2.2/Conv""
How to modify the reg_layer and cls_layer to correspond to the right node name in my model?

2

Hey @sxk0988

To modify the reg_layer and cls_layer to correspond to your model’s output node names, replace the values with the names from your model as seen in Netron. For example:

reg_layer:
  - /model.22/cv2.0/cv2.0.2/Conv
  - /model.22/cv2.1/cv2.1.2/Conv
  - /model.22/cv2.2/cv2.2.2/Conv

cls_layer:
  - /model.22/cv3.0/cv3.0.2/Conv
  - /model.22/cv3.1/cv3.1.2/Conv
  - /model.22/cv3.2/cv3.2.2/Conv

Ensure these names exactly match the nodes in your model.

3

Hello @omria ,
Thanks for your help. I used the dfc-studio to check the ONNX model and parsing the model to HAR file. From the dfc-studio, I can find the right node name for nms_config.json to fix this problem. Thanks for your help.

4

Hi, I saw an error during optimization “hailo_sdk_client.sdk_backend.sdk_backend_exceptions.AllocatorScriptParserException: Cannot infer bbox conv layers automatically. Please specify the bbox layer in the json configuration file.”. I guess it may be related to this file. Where can I find this file? I don’t see it in the Dataflow compiler package. Thanks!

5

Hello @TT2024,
I used the jason configuration file from below link. Maybe you can try to find your one.

6

Thanks for the info!