dinov2

dinov2 copied to clipboard

Hi, I've identified a potential issue within the data augmentation pipeline of DINOv2 that appears to deviate from the expected behavior observed in DINO and MoCo-v3. According to the source...

Separius

Efficiently handling large-scale vision datasets

2

Hi, I'm using DINOv2 to pretrain a ViT on a dataset significantly larger than ImageNet22k (between 100M and 1B jpg images). I sticked to the ImageNet22k dataset class for handling...

clemsgrs

Cannot create a conda environment.

6

Hi! I have some issues with creating the environment. Both mamba and conda return error. While running `conda env create -f conda.yaml` I encounter the error: ``` Could not solve...

LiubovShilova

Python version?

9

It is not clear what Python version is suitable for this project. I am unable to install the required dependencies when using versions 3.6, 3.8, 3.9, 3.10 and 3.11 python3.11:...

jamesheatonrdm

ViT-S and drop_path_rate

3

Hi, I'm trying to train DINOv2 with the ViT-S backbone. I've noticed that the model training crashes if `drop_path_rate` > 0. In particular, I get the following error ``` RuntimeError:...

vladchimescu

[request] Additional model exports (ONNX, CoreML, ...)

2

Related: - #102 - #129 - #19 - #191 - #206

patricklabatut

[draft]Add onnx export script

5

## Objective Export dinov2 to onnx for serving. ## Changed - Fix hubconf to get `Attention` to export onnx - Fix minor part in model - Add onnx export script...

RRoundTable

Retrieving original pixel data from patches.

6

I am trying to retrieve the pixel data in relation to features rather than patches in relation to features. My issue is that I can not seem to find a...

CaedenMotley

Hello DINO-V2 Community, I have a dataset for monocular depth estimation, and I am planning to fine-tune pre-trained ViT backbone + DPT head on this dataset. While fine-tuning it, backbone...

GoktugGuvercin

Do feature vectors for DINOv2 include small objects?

5

Hi, When I visualize the features via PCA I'm able to see small objects but I'm not sure if this means the 1024 feature vector or ViT-L from DINOv2 must...

smandava98