app.py

import gradio as gr
import torch
import torchvision.transforms as T

import numpy as np
from PIL import Image

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# DINOv2

# Select checkpoint
dinov2_ckpt = ['dinov2_vits14', 'dinov2_vitb14', 'dinov2_vitl14', 'dinov2_vitg14'][1]
dinov2 = torch.hub.load('facebookresearch/dinov2', dinov2_ckpt)

dinov2.to(device)
print()

transform_image = T.Compose([
    T.Resize((224, 224)),
    T.ToTensor(),
    T.Normalize(mean=[0.485, 0.456, 0.406],
                        std=[0.229, 0.224, 0.225])
])

def predict(image):
  """
  Predict the identity of an image.

  Args:
    image: A PIL Image object.

  Returns:
    A string representing the predicted identity of the image.
  """

  # Convert the image to a tensor.
  transformed_img = transform_image(image)[:3].unsqueeze(0).to(device)

  # Get the embedding of the image.
  with torch.no_grad():
    embedding = dinov2(transformed_img)
    print(embedding.shape)
    embedding = embedding[0].cpu().numpy().tolist()
    print(embedding)
  return {
      "embedding": embedding
  }

# Create a Gradio interface.
interface = gr.Interface(
    fn=predict,
    inputs=[gr.Image(type='pil')],
    outputs=[gr.JSON()],
    title="DINOv2 Embedding",
    description=dinov2_ckpt
)

# Start the Gradio server.
interface.launch()