Missing config : num_masks

[lhkyyh]

When I load my trained model, I encounter this problem：Missing config : num_masks

[CVHub520]

Hey there! @lhkyyh, 😄

It seems like you are encountering an issue that raises an error: Missing config: num_masks.

Perhaps you can resolve this by replacing your local YAML file with the following configuration:

type: yolov8_seg
name: yolov8hz
display_name: yolov8hz
model_path: hz.onnx
nms_threshold: 0.20
confidence_threshold: 0.20
classes:
  - '0'
  - '1'

Note:

• Label names starting with numbers should be enclosed in single quotes (' ').
• The anchors and stride fields are only allowed to be defined in the configuration file.

Finally, you can inspect the input and output nodes of your ONNX file using Netron and paste the details here.

[lhkyyh]

Thanks for the answer, the same issue occurs even though I followed the config file above, this is the input and output nodes of my ONNX file

[lhkyyh]

input
images
name: images
tensor: float32[1,3,640,640]
output
output0
name: output0
tensor: float32[1,38,8400]
output1
name: output1
tensor: float32[1,32,160,160]

[CVHub520]

Thank you for your response. If you're still encountering issues, it's often helpful to try cloning the latest source code, as this can resolve many unknown errors that might be present in older versions.

After you've cloned and run the latest source code, if the problem persists, please prepare a compressed package including the ONNX file, the YAML configuration file, and an image that you're certain should be detectable by the model. You can then send this package to the following email address for further assistance: [email protected].

Thank you for your cooperation!

[CVHub520]

After testing your sending file, it seems that everything is normal. As memtion above, please using the latest source code via git clone https://github.com/CVHub520/X-AnyLabeling.git, and make sure your GUI is looks like below:

Btw, an optimal piece of advice would be to set the classes as defined in your ONNX file, rather than using ‘0’ or ‘1’:

type: yolov8_seg
name: yolov8hz
display_name: yolov8hz
model_path: hz.onnx
nms_threshold: 0.20
confidence_threshold: 0.20
classes:
  - crack
  - blocky crack

[lhkyyh]

Thanks for your help, it maybe my computer's problem , It runs well in other computer, good lucky to all of you

[CVHub520]

Thank you for the update. I believe I might have identified a potential cause for the issue on your computer. It's possible that your environment was contaminated, perhaps by accidentally installing AnyLabeling. This can be resolved by running pip uninstall anylabeling.

For the best practice, I recommend creating a new Conda environment and then reinstalling the dependencies with pip install -r requirements.txt. This ensures a clean setup and can help avoid such issues in the future.

I'm glad to hear that it's running well on another computer. Wishing you all the best too!

[lhkyyh]

I get it，I create a new environment and replace a file named .anylabelingrc in my computer, it runs well now, Thanks so much again!