Embedding Atlas is an open-source tool from Apple that provides interactive visualizations for large-scale embedding spaces. It lets you explore, cross-filter, and search embeddings alongside their metadata, making it easier to understand the structure and distribution of high-dimensional data.
Visualization & Exploration
- Automatic data clustering and labeling to interactively visualize and navigate overall data structure
- Kernel density estimation and density contours for easily distinguishing dense regions from outliers
- Order-independent transparency ensuring clear, accurate rendering of overlapping data points
- Smooth performance at scale, handling up to a few million points through a modern rendering stack
Search & Analysis
- Real-time search and nearest-neighbor lookup to find data similar to a given query or existing point
- Multi-coordinated views for metadata exploration, linking and filtering data across columns interactively
- Cross-filtering between embedding space and metadata dimensions for deeper analytical insights
Rendering & Performance
- WebGPU implementation with WebGL 2 fallback for broad browser compatibility
- GPU-accelerated rendering pipeline delivering fast, fluid interaction even on large datasets
- Lightweight front-end designed for exploratory data analysis without server-side dependencies