lightly-ai/lightly-studio

What it solves

LightlyStudio is a data curation and management tool designed to handle the complexities of preparing large-scale image and video datasets for AI training. It solves the problem of manually sorting through millions of files to find relevant samples, remove duplicates, and manage annotations efficiently without needing to upload data to a cloud service.

How it works

The tool runs as a local server that opens in a web browser. Users index their local images or videos (including those in cloud storage like S3, GCS, or Azure) into a local database. It uses embeddings to visualize the relationship between images, allowing users to lasso-select clusters of similar images to filter the grid. It also provides a Python API for scriptable indexing, querying, and exporting data in formats like COCO and YOLO.

Who it’s for

It is primarily for computer vision engineers and data scientists who need to curate, label, and evaluate model predictions against ground truth for image and video datasets.

Highlights

• Local-first processing: Data and images never leave the machine, ensuring privacy and security.
• High performance: Capable of handling 2M+ images and embeddings on a single MacBook M1.
• Intelligent curation: Uses embedding plots and sampling techniques to pick diverse and typical samples to reduce labeling costs.
• Extensible: Supports plugins for auto-labeling, such as the SAM (Segment Anything Model) plugin.
• Comprehensive workflow: Covers the entire pipeline from raw data indexing to model evaluation and exporting for training.