Dataset Viewer
The dataset viewer is not available for this subset.
Job has been terminated due to a temporary spike in resource usage and may be restarted later.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Section: Hierarchical Concept Segmentation Benchmark Suite

We curate a hierarchical concept segmentation benchmark suite to evaluate the out-of-distribution generalization of ConceptSeg-R1 across three cognitive levels: Context-Independent (CI) , Context-Dependent (CD) , and Context-Reasoning (CR) concepts. This suite integrates existing datasets into a unified taxonomy, supporting both rule induction training and structured evaluation. The following table summarizes the full benchmark composition.

Concept Level Directory Name Source Paper/Dataset Name Description
Context-Independent (CI) coco2014_Living COCO One-shot learning for semantic segmentation. Common living objects (e.g., people, animals).
Context-Independent (CI) coco2014_Artifact COCO One-shot learning for semantic segmentation. Man-made artifact categories (e.g., vehicles, tools).
Context-Independent (CI) ultra_rare iNaturalist Sam 3: Segment anything with concepts. Long-tail classes (bottom 1% of model confidence).
Context-Reasoning (CI) rare iNaturalist Sam 3: Segment anything with concepts. Evaluation of OOD / Rare classes.
Context-Reasoning (CI) fewshot1000 FSS-1000 FSS-1000: A 1000-Class Dataset for Few-shot Segmentation Consistency: Identifying shared patterns across support sets.
Context-Reasoning (CI) CoSOD3k1024 COSOD3K Co-Salient Object Detection: A Benchmark and Algorithms Co-saliency: Reasoning to find co-existing objects.
Context-Dependent (CD) DUTS DUTS DUTS: A Large-scale Dataset for Salient Object Detection Saliency: Targets that stand out from the background.
Context-Dependent (CD) COD10K1024 COD10K COD10K: A Large-scale Camouflaged Object Detection Dataset Camouflage: Targets blending into surroundings.
Context-Dependent (CD) transparent1024 Trans10K Trans10K: A Large-scale Dataset for Transparent Object Segmentation Transparency: Optical refraction through materials.
Context-Dependent (CD) Shadow_detection SBU Large-scale training of shadow detectors with noisily-annotated shadow examples. Shadows: Alterations by contextual interactions.
Context-Dependent (CD) ESDIDefects ESDIs-SOD Autocorrelation aware aggregation network for salient object detection of strip steel surface defects Industrial Anomaly: Manufacturing surface defects.
Context-Dependent (CD) Polyp Kvasir/CVC Pranet: Parallel reverse attention network for polyp segmentation. Medical: Colon Polyp lesions identification.
Context-Dependent (CD) Breast_Tumor Dataset-B . Dataset of breast ultrasound images Medical: Breast Ultrasound tumors identification.
Context-Dependent (CD) isic2018 ISIC2018 Skin lesion analysis toward melanoma detection 2018: A challenge hosted by the international skin imaging collaboration (isic) Medical: Skin lesion analysis relative to healthy tissue.
Context-Reasoning (CR) MGrounding-630k MGrounding Multi-Image Grounding for Visual Reasoning Logic: Reasoning to identify distinct or shared objects.
Context-Reasoning (CR) MIG-Bench MIG Migician: Revealing the magic of free-form multi-image grounding in multimodal large language models Complex CR: Includes Spatio-temporality and View Difference.
Downloads last month
260