DINOv2: Learning Robust Visual Features without Supervision

Self-supervised vision pretraining produced features that were brittle across distributions and needed finetuning per task.

Combines teacher-student self-distillation with an automatic curated-data pipeline, then distils a 1B-parameter ViT into a series of smaller ViTs that become the released backbone. No language supervision, so the encoder is a pure image feature extractor.

Published in Transactions on Machine Learning Research (openalex shows the TMLR ISSN 2835-8856 for this title); the schema's venue enum has no TMLR entry, so 'other' is the honest encoding. DINOv2 vision backbones are widely embedded in production vision systems and as open-vocabulary detectors, usually alongside CLIP text rather than instead of it. The distillation cascade matters for serving: the small distilled variants are what ships, and inference cost is set by the patch grid at the deployed input resolution. VENUE CORRECTION + VENUE_TRACK 2026-10-04 (paper-fields pass, agent w3p): venue 'other' -> 'iclr', venue_track 'main'. EVIDENCE: the STATIC proceedings index for this venue lists this exact title. Venue 'ICLR' paper page: ICLR 2025 poster 31510. The listing was corroborated by author overlap, not by title alone, because these titles are duplicated across venues and years and a title-only match is exactly the kind of thing that produced this repo's two previously-shipped wrong arXiv IDs. BOTH VALUES RETAINED PER THE REPO CONTRACT: the prior value was 'other'. status set to contested because this changes a peer-reviewed attribution. arxiv_id unchanged (2304.07193). Method note: OpenReview's api2 hard-429s under load and pre-2024 venues are unreachable through it, so these were checked against static proceedings indexes (iclr.cc/virtual/<year>/papers.html and neurips.cc/virtual/<year>/papers.html, with the per-paper JSON-LD creditText field as the venue-year assertion). YEAR FIELD CORRECTED 2026-10-04 (same pass as the venue correction above): year 2023 -> 2025 to match the ICLR 2025 venue year. The prior value 2023 is the arXiv posting year and remains true of the preprint. status stays contested. This drift is why the venue correction and the year correction were done together rather than separately: setting venue without setting year leaves the record internally inconsistent. 2026-10-10: ICLR 2025 accepted titles index (idx3.py: https://iclr.cc/virtual/2025/papers.html?filter=titles) and poster page (final1.py: https://iclr.cc/virtual/2025/poster/31510) confirm "DINOv2: Learning Robust Visual Features without Supervision". Source pap-mm-dinov2 notes cite TMLR (ISSN 2835-8856) while ICLR 2025 index confirms ICLR 2025. Sources disagree on venue; status remains contested.