DINOv2: Learning Robust Visual Features without Supervision
Abstrak
The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques to scale our pretraining in terms of data and model size. Most of the technical contributions aim at accelerating and stabilizing the training at scale. In terms of data, we propose an automatic pipeline to build a dedicated, diverse, and curated image dataset instead of uncurated data, as typically done in the self-supervised literature. In terms of models, we train a ViT model (Dosovitskiy et al., 2020) with 1B parameters and distill it into a series of smaller models that surpass the best available all-purpose features, OpenCLIP (Ilharco et al., 2021) on most of the benchmarks at image and pixel levels.
Topik & Kata Kunci
Penulis (26)
M. Oquab
Timothée Darcet
Théo Moutakanni
Huy V. Vo
Marc Szafraniec
Vasil Khalidov
Pierre Fernandez
Daniel Haziza
Francisco Massa
Alaaeldin El-Nouby
Mahmoud Assran
Nicolas Ballas
Wojciech Galuba
Russ Howes
Po-Yao (Bernie) Huang
Shang-Wen Li
Ishan Misra
Michael G. Rabbat
Vasu Sharma
Gabriel Synnaeve
Hu Xu
H. Jégou
J. Mairal
Patrick Labatut
Armand Joulin
Piotr Bojanowski
Akses Cepat
- Tahun Terbit
- 2023
- Bahasa
- en
- Total Sitasi
- 7509×
- Sumber Database
- Semantic Scholar
- DOI
- 10.48550/arXiv.2304.07193
- Akses
- Open Access ✓