avrecum/

coco_bbox_quality_sam3_masks

A dataset of images containing multiple same-class objects, each annotated with discriminative referring expressions and SAM-generated segmentation masks. Total images: 925 Parquet shards: 2

Each parquet row represents one image with all its annotated objects.

Column Type Description

image binary Raw image bytes (JPEG)

image_id string Unique image identifier

image_width int32 Image width in pixels

image_height… See the full description on the dataset page: https://huggingface.co/datasets/avrecum/coco_bbox_quality_sam3_masks.