[IDEA] Detect the damage zone in the picture

[llermaly]

Using LLM to detect the damage zones is being very hard, but we can use a library like this one:

https://github.com/facebookresearch/detectron2

They have a dataset of damaged cars so probably is straightforward to mark the damaged zone:

https://www.kaggle.com/code/lplenka/detectron2-car-damaged-parts-detection

2 decisions to make here:

1. Do we use LLM or detectron
2. Do we implement this for the first hackathon or for the second one (probably the second one is best)

[llermaly]

Given the Clarifai challenge we can use their damage detection model from here:

https://www.kaggle.com/datasets/lplenka/coco-car-damage-detection-dataset

Using this loader:

https://github.com/Clarifai/clarifai-python/tree/master/clarifai/datasets/upload/loaders