dinov2-I16

Source code on GitHub.

README

DINOv2 patch

This is a patch for the original repository to make it work with the latest version of PyTorch (>2.1).

Install the dependencies using the following command:

``bash conda env create -f conda.yaml conda activate dinov2 `

Then run the following command to install the package:

`bash pip install -e . `

Then add the training images to data/train.

You can now start the training using torchrun instead of submitit. The following command will start the training on 2 GPUs:

`bash torchrun --nproc_per_node 2 train.py --config-file dinov2/configs/train/vitl16_short_custom.yaml --output-dir output `

Thanks a lot to https://github.com/csaroff/dinov2 for an example of custom dataset.

:new: [2023-10-26] Added DINOv2 backbones with registers, following Vision Transformers Need Registers.

DINOv2: Learning Robust Visual Features without Supervision

Meta AI Research, FAIR

Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V. Vo, Marc Szafraniec, Vasil Khalidov, Patrick Labatut, Armand Joulin, Piotr Bojanowski

[Paper #1] Paper #2] [Blog] [Demo] [BibTeX]

PyTorch implementation and pretrained models for DINOv2. For details, see the papers: DINOv2: Learning Robust Visual Features without Supervision and Vision Transformers Need Registers.

DINOv2 models produce high-performance visual features that can be directly employed with classifiers as simple as linear layers on a variety of computer vision tasks; these visual features are robust and perform well across domains without any requirement for fine-tuning. The models were pretrained on a dataset of 142 M images without using any labels or annotations.

https://github.com/facebookresearch/dinov2/assets/60359573/f168823e-7922-415a-b429-578badf5c356

Pretrained models