Spot outliers by lassoing an embedding map | Voxel51

Embeddings turn a dataset into a map. Harpreet Sahota, hacker-in-residence at Voxel51, shows how similar images group into islands and the points far from the rest stand out. Lasso a region of the embedding map and the matching samples open beside it, so you inspect the outliers first instead of scrolling through thousands of images. Cut from Curating Robotic Data Collection with Voxel51: https://www.youtube.com/watch?v=VH9D9ctTmAQ 🔗 Learn more Outlier detection: https://voxel51.com/glossary/outlier-detection Embedding space: https://voxel51.com/glossary/embedding-space Data curation: https://voxel51.com/glossary/data-curation Get started free: https://link.voxel51.com/yt-harpreet-2 💬 Join the community: https://community.voxel51.com Voxel51 is the visual AI data platform for physical AI. It covers data curation, data annotation, and model evaluation for visual AI and physical AI across images, video, 3D, and multimodal sensor data. #Voxel51 #Embeddings #PhysicalAI