Image Classification and Prediction with DINOv2 in Google Colab

NOTE: This is validated in Google Colab T4 GPU

Here we are going to discuss how to classify images using DINOv2 embeddings and a C-Support Vector Classification (SVC) linear classification model. We will have a model trained on the Kaggle Vegiable Dataset that classifies vegitable images.

Clone DINOv2 install dependancies

Step 1

git clone https://github.com/facebookresearch/dinov2.git

Step 2

cd dinov2

Step 3

pip3 install -r requirements.txt

For Download Kaggle Vegiable Dataset follow below steps

Step1

pip3 install kaggle

Step 2

• Goto Kaggle Settings
• Click on Create New Token will get kaggle.json file
• Put this json filr to ~/.kaggle/

Step 3

Run following command that will download respecctive dataset

kaggle datasets download -d misrakahmed/vegetable-image-dataset

Execute Notebook

dinov2_classify_and_predict.ipynb

Refference

1. https://github.com/facebookresearch/dinov2
2. https://www.kaggle.com/datasets/misrakahmed/vegetable-image-dataset
3. https://blog.roboflow.com/how-to-classify-images-with-dinov2/