Clarification on Results for Four Sections(normal, degrading):(normal, emergency vehicles)

[gaya1thri]

Sorry for the inconvenience, we have code for cases background+ego and Background+ego+rescue, can you please provided the codes for all the 4 cases? Thank you so much.

[gaya1thri]

I appreciate your efforts in developing this repository. I would like to understand how the results for the four sections are generated and structured. Specifically:
How are the final outputs interpreted?

Your guidance would be greatly appreciated. Looking forward to your response.

[pangay]

Thanks for your message and for your interest in the work.
I’m not entirely sure how the model has been trained in your setup, or whether an LLM was used for scene understanding. Based on the current results, I’d suggest first checking the rescue case separately to see if there is any noticeable improvement.

After that, please take a look at the logs to verify whether the LLM correctly interprets the rescue information. This should help clarify where the issue may lie.