ImageNet-S
Emerging4papers using it
2025first seen
'ImageNet-S' is a dataset that contains pixel-level silhouettes of geometrically distinct classes used to evaluate the geometric comprehension capabilities of Vision-Language Models (VLMs).
'ImageNet-S' is a dataset that contains pixel-level silhouettes of geometrically distinct classes used to evaluate the geometric comprehension capabilities of Vision-Language Models (VLMs).