// ◐ World Models & Self-Supervised Learning
DINOv2 (Self-Supervised Visual Features)
Label-free self-distillation on curated images yields general-purpose visual features.
DINOv2 自监督视觉特征 aka DINO 2023 Intermediate core
IN PLAIN WORDS
A student and a teacher network see different crops of the same image; the student predicts the teacher’s output, and the teacher is a moving average of the student. Trained on 142M curated images, DINOv2 features work off the shelf – no fine-tuning – for classification, segmentation, depth and retrieval.
Used in: Visual encoder in Depth Anything and others; a frozen backbone for retrieval and segmentation
TIMELINE · prerequisites → this → what builds on it
01Original paper
DINOv2: Learning Robust Visual Features without Supervision
Maxime Oquab, Timothée Darcet, Théo Moutakanni, et al. · 2023
arXiv ↗ ✓ Verified · arXiv API
02Where it sits
Learn the left column first; the right column builds on this one.
Learn first
Then you can learn
- Nothing in the atlas builds on it yet
Related:Masked Autoencoders (MAE)CLIP (Contrastive Language–Image Pretraining)Joint-Embedding Predictive Architecture (JEPA)
Left: prerequisites · right: builds on this · dashed: related. Full map →
03Learning resources
Official docs, explainers, courses and reference code. All links checked.
04Where the course teaches it
For AI / data
- Verification: verified, checked 2026-10-22.
- JSON: /api/learn/algorithms.json (id:
dinov2)