Module dinov2

Module dinov2 

Source
Expand description

Implementation of the DINOv2 models from Meta Research.

This module implements the DINOv2 vision transformer model from Meta AI Research. DINOv2 is a self-supervised learning model that can learn visual features without using any labeled data. See: “DINOv2: Learning Robust Visual Features without Supervision”

§Running an example with color map and CUDA

cargo run \
  --features cuda,depth_anything_v2 \
  --package candle-examples \
  --example depth_anything_v2 \
  -- --color-map \
  --image candle-examples/examples/yolo-v8/assets/bike.jpg

§Running as an ImageNet classifier

The model returns the probability for the image to belong to each of the 1000 ImageNet categories.

cargo run \
  --example dinov2 \
  --release \
  -- --image candle-examples/examples/yolo-v8/assets/bike.jpg

> mountain bike, all-terrain bike, off-roader: 43.67%
> bicycle-built-for-two, tandem bicycle, tandem: 33.20%
> crash helmet            : 13.23%
> unicycle, monocycle     : 2.44%
> maillot                 : 2.42%

Structs§

DinoVisionTransformer

Functions§

vit_small