Lightly

Integrated, open-source-at-its-core multimodal/computer-vision data platform for curation, annotation, QA, evaluation, collaboration, dataset versioning, embeddings, and export, with GUI, Python SDK, and API.

Computer-vision training toolkit covering self-supervised pretraining, pseudo-labeling/autolabeling, distillation, and fine-tuning for detection, segmentation, and classification models, including DINOv2/v3 workflows.

Open-source PyTorch-based self-supervised learning framework for training vision models without labels; supports modular losses and heads, custom backbones, distributed training, and many SSL methods.

Expert AI training-data services for LLMs, AI agents, computer vision, multimodal models, RL environments, RLHF, model evaluation, synthetic data, and prompt generation.