Computer Vision & Multimedia Datasets

These repositories specialize in images, video, annotations, and multimodal resources. Ideal for training models in object detection, segmentation, captioning, and scene understanding.

Tools:

• Visual Genome – Image dataset with region descriptions, relationships, and attributes for scene understanding. Commonly used in VQA and multimodal AI.

• Million Song Dataset – A comprehensive music dataset for recommendation systems, beat analysis, and music intelligence.

• Tabby Vision Dataset – Offers labeled images for vision research, especially in medical or environmental contexts.