MAOAM: Unified Object and Material Selection with Vision-Language Models

MAOAM (Mask Any Object And Material) is a unified selection framework for precise object and material-level selection across text- and click-based interactions. This repository contains a 10% subset of the material annotations from the associated paper, featuring per-region text descriptions and VQA questions across three sets: SynMat, RealMat, and SAMa. The dataset was authored by jpark677 and last updated on Hugging Face in June 2026.

Use Cases

• Train vision-language models for interactive segmentation based on the described text- and click-based interactions.
• Benchmark material recognition and segmentation models using the provided SynMat, RealMat, and SAMa annotation sets.
• Generate or evaluate visual question answering (VQA) models based on the per-region VQA questions included in the data.

Strengths

• The dataset is derived from a published research paper with a defined framework (MAOAM).
• It contains a 10% subset of material annotations with per-region text descriptions and VQA questions.
• Annotations are organized across three distinct sets: SynMat, RealMat, and SAMa.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Hugging Face dataset authored by jpark677, associated with the MAOAM research paper.

Collection Method

Likely contains annotations generated for the MAOAM research framework.

Freshness

Last updated 2026-06-06 03:31:45; freshness should be verified.