AssertionError in Detectron2 PanopticFPN with Custom Mapper #5411
Labels
invalid/unrelated
unrelated to this project or invalid type of issues
needs-more-info
More info is needed to complete the issue
I'm encountering an AssertionError: assert "sem_seg" in batched_inputs[0] while training a Detectron2 model with PanopticFPN and a custom mapper for data augmentation.
Custom Mapper:
Briefly explain the purpose of your custom mapper (e.g., increase data augmentation).
Include the relevant code snippet for custom_mapper. Remove sensitive information if necessary. Consider linking to a Gist or repository if the code is lengthy.
Full Error Traceback:
Provide a complete copy of the error traceback, including line numbers and any associated error messages. You can paste this into a code block using triple backticks (`````) on GitHub.
Expected Behavior:
Describe how you expected the code to work (e.g., model to train with augmentations applied).
Troubleshooting Steps:
Mention any steps you've already tried to resolve the issue (e.g., reviewed similar examples, double-checked code for assumptions about data format).
Environment:
List key components (OS, Python version, Detectron2 version, GPU details).
Additional Context (Optional):
Include any relevant information not covered above, such as configuration files or specific datasets.
If you have specific questions or areas where you need guidance, feel free to mention them.
Example Code Snippet:
Python
def custom_mapper(dataset_dict):
# ... (Your custom data augmentation logic here) ...
return dataset_dict
Additional Considerations:
Double-check potential typos or errors in your custom_mapper function.
Verify if any assumptions about input data formats or image dimensions might be causing problems.
Consider simplifying your custom mapping logic temporarily to isolate the root cause.
Review Detectron2 documentation or search for similar issues in the community forums.
The text was updated successfully, but these errors were encountered: