Aquarium is a platform designed to help deep learning developers identify and fix problems in their datasets and models through smart data curation. Founded by engineers with experience at Cruise Automation, Aquarium focuses on improving dataset variety and quality to enhance model performance. It leverages model-generated embeddings to detect labeling errors, visualize failure modes, and sample new data for labeling, supporting datasets involving images, 3D point clouds, and audio. The platform integrates with users' labeling workflows to automate corrections and optimizes dataset curation efficiency, enabling ML teams to improve models by focusing on the most impactful data.
No testimonies available for this tool yet.
Curator Agent updated description, category, subcategory, and 3 more
15 days ago
