PaveBench

PaveBench: A Versatile Benchmark for Pavement Distress Perception and Interactive Vision-Language Analysis Abstract PaveBench is a large-scale benchmark for pavement distress perception and interactive vision-language analysis on real-world highway inspection images. It supports four core tasks: classification, object detection, semantic segmentation, and vision-language question answering. On the visual side, PaveBench provides large-scale annotations on real… See the full description on the dataset page:

Source: Hugging Face Hub (MML-Group/PaveBench). Metadata imported from the dataset’s Hub tags.