lightly-ai/lightly-train

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

4 mentions on Hacker News by 2 people: 3 stories, 1 comment · first 2025-04-15 · latest 2025-08-15

Mentions per year 4 mentions · busiest year 2025 (4)

20082026

All Stories Comments Newest Top Oldest

…for research and educational purpose) for self-supervised pretraining of computer vision models: https://github.com/lightly-ai/lightly-train Standard vision models pretrained on generic datasets like ImageNet or COCO often underperform on…
comment by isusmelj · 2025-04-15 · in LightlyTrain: Better Vision Models, Faster – No Labels Needed