ADE20K

The ADE-20K dataset is a comprehensive benchmark that contains a diverse set of images annotated with pixel-level semantic segmentation, used to evaluate the performance of models in understanding and interpreting complex scenes.

Papers using ADE20K (8)

A novel Framework for Open-Vocabulary Multi-Object Recognition using CLIP2026SpatialBoost: Enhancing Visual Representation through Language-Guided Reasoning2026LangHOPS: Language Grounded Hierarchical Open-Vocabulary Part Segmentation2025Scene-aware Urban Design: A Human-ai Recommendation Framework Using Co-occurrence Embeddings And Vision-language Models2025Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks2022 · 151 citesONE-PEACE: Exploring One General Representation Model Toward Unlimited Modalities2023 · 42 citesMTA-CLIP: Language-Guided Semantic Segmentation with Mask-Text Alignment2024 · 2 citesEmergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models2023