Paper: GRAID: Enhancing Spatial Reasoning of VLMs Through High-Fidelity Data Generation
Project Page
This dataset was generated using GRAID (Generating Reasoning questions from Analysis of Images via Discriminative artificial intelligence), a framework for creating spatial reasoning datasets from object detection annotations, as presented in the linked paper.
GRAID transforms raw object detection data into… See the full description on the dataset page:
https://huggingface.co/datasets/kd7/graid-waymo-unique-wd.
