Dpt Dinov2 Large Kitti

🚀 Model Card: DPT model with DINOv2 backbone

This model showcases a powerful depth estimator by using the DPT framework with DINOv2 as the backbone.

🚀 Quick Start

Model Details

The DPT (Dense Prediction Transformer) model with a DINOv2 backbone, as proposed in DINOv2: Learning Robust Visual Features without Supervision by Oquab et al.

DPT architecture. Taken from the original paper.

Resources

• DINOv2 Paper
• DPT Paper

Use with Transformers

from transformers import AutoImageProcessor, DPTForDepthEstimation
import torch
import numpy as np
from PIL import Image
import requests

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)

image_processor = AutoImageProcessor.from_pretrained("facebook/dpt-dinov2-large-kitti")
model = DPTForDepthEstimation.from_pretrained("facebook/dpt-dinov2-large-kitti")

# prepare image for the model
inputs = image_processor(images=image, return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)
    predicted_depth = outputs.predicted_depth

# interpolate to original size
prediction = torch.nn.functional.interpolate(
    predicted_depth.unsqueeze(1),
    size=image.size[::-1],
    mode="bicubic",
    align_corners=False,
)

# visualize the prediction
output = prediction.squeeze().cpu().numpy()
formatted = (output * 255 / np.max(output)).astype("uint8")
depth = Image.fromarray(formatted)

💻 Usage Examples

Basic Usage

from transformers import AutoImageProcessor, DPTForDepthEstimation
import torch
import numpy as np
from PIL import Image
import requests

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)

image_processor = AutoImageProcessor.from_pretrained("facebook/dpt-dinov2-large-kitti")
model = DPTForDepthEstimation.from_pretrained("facebook/dpt-dinov2-large-kitti")

# prepare image for the model
inputs = image_processor(images=image, return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)
    predicted_depth = outputs.predicted_depth

# interpolate to original size
prediction = torch.nn.functional.interpolate(
    predicted_depth.unsqueeze(1),
    size=image.size[::-1],
    mode="bicubic",
    align_corners=False,
)

# visualize the prediction
output = prediction.squeeze().cpu().numpy()
formatted = (output * 255 / np.max(output)).astype("uint8")
depth = Image.fromarray(formatted)

📚 Documentation

Model Use

Intended Use

The model is intended to showcase that using the DPT framework with DINOv2 as backbone yields a powerful depth estimator.

BibTeX entry and citation info

@misc{oquab2023dinov2,
      title={DINOv2: Learning Robust Visual Features without Supervision}, 
      author={Maxime Oquab and Timothée Darcet and Théo Moutakanni and Huy Vo and Marc Szafraniec and Vasil Khalidov and Pierre Fernandez and Daniel Haziza and Francisco Massa and Alaaeldin El-Nouby and Mahmoud Assran and Nicolas Ballas and Wojciech Galuba and Russell Howes and Po-Yao Huang and Shang-Wen Li and Ishan Misra and Michael Rabbat and Vasu Sharma and Gabriel Synnaeve and Hu Xu and Hervé Jegou and Julien Mairal and Patrick Labatut and Armand Joulin and Piotr Bojanowski},
      year={2023},
      eprint={2304.07193},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

📄 License

This project is licensed under the Apache-2.0 license.