We need to build four different datasets. For car damage detection, we will think about all the possible inputs. It could be a car in good condition, or a car with different damage levels, or it could be an unrelated image of a car.
We will do the same as shown in the following screenshots:
Here is the dataset to identify heavily damaged cars:
googleimagesdownload -k "heavily damaged car" -sk 'red,blue,white,black,green,brown,pink,yellow' -l 500
Here are some sample pictures that were captured for heavily damaged cars that are red:
Here are some sample pictures that were captured for heavily damaged cars that are blue: ...
