lightly-ai/lightly-train

All-in-one training for vision models (YOLO, ViTs, RT-DETR, DINOv3): pretraining, fine-tuning, distillation.

30 best points · 11 comments · 3 Hacker News posts · 1,727 GitHub stars · View on GitHub

Hacker News discussions

• LightlyTrain: Better Vision Models, Faster – No Labels Needed (30 points, 11 comments, by michal-l, 2025-04-15)
• Show HN: LightlyTrain – Pretrain YOLO/ResNet on unlabeled data beats ImageNet (3 points, 0 comments, by isusmelj, 2025-04-15)
• Show HN: Distill DINOv3 into your own model (1 points, 0 comments, by isusmelj, 2025-08-15)