Distill Any Depth Small Hf

🚀 轻量级深度蒸馏模型 - Transformers版本

Distill-Any-Depth是一种全新的单目深度估计模型，它采用了我们提出的知识蒸馏算法进行训练，达到了当前最优水平（SOTA）。该模型在论文 Distill Any Depth: Distillation Creates a Stronger Monocular Depth Estimator 中被首次提出。此模型的检查点与transformers库兼容，你可以通过 在线演示 体验其效果。

🚀 快速开始

以下是使用该模型进行零样本深度估计的方法：

基础用法

from transformers import pipeline
from PIL import Image
import requests
# load pipe
pipe = pipeline(task="depth-estimation", model="xingyang1/Distill-Any-Depth-Small-hf")
# load image
url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
image = Image.open(requests.get(url, stream=True).raw)
# inference
depth = pipe(image)["depth"]

高级用法

from transformers import AutoImageProcessor, AutoModelForDepthEstimation
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("xingyang1/Distill-Any-Depth-Small-hf")
model = AutoModelForDepthEstimation.from_pretrained("xingyang1/Distill-Any-Depth-Small-hf")

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

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

# interpolate to original size and visualize the prediction
post_processed_output = image_processor.post_process_depth_estimation(
    outputs,
    target_sizes=[(image.height, image.width)],
)

predicted_depth = post_processed_output[0]["predicted_depth"]
depth = (predicted_depth - predicted_depth.min()) / (predicted_depth.max() - predicted_depth.min())
depth = depth.detach().cpu().numpy() * 255
depth = Image.fromarray(depth.astype("uint8"))

📄 许可证

本项目采用MIT许可证。

📚 详细文档

如果你觉得这个项目有用，请考虑引用以下文献：

@article{he2025distill,
  title   = {Distill Any Depth: Distillation Creates a Stronger Monocular Depth Estimator},
  author  = {Xiankang He and Dongyan Guo and Hongji Li and Ruibo Li and Ying Cui and Chi Zhang},
  year    = {2025},
  journal = {arXiv preprint arXiv: 2502.19204}
}

👨‍💻 模型卡片作者

Parteek Kamboj

📋 模型信息