VisQuant: A Synthetic Benchmark for Object Counting and Spatial Reasoning
收藏数据链接：
官方服务：
资源简介：
VisQuant is a synthetic dataset of 100 annotated image scenarios, purpose-built to evaluate AI systems on object counting, spatial layout understanding, and visual question answering (VQA).This dataset is ideal for benchmarking vision-language models (e.g. GPT-4V, Claude, Gemini), and for training reasoning agents that must understand objects in relation to one another in structured scenes.
提供机构：
IEEE DataPort创建时间：
2025-04-06
