README.md
4.3 KB · 109 lines · markdown Raw
1 ---
2 library_name: transformers
3 library: transformers
4 license: cc-by-nc-4.0
5 tags:
6 - depth
7 - relative depth
8 pipeline_tag: depth-estimation
9 widget:
10 - inference: false
11 ---
12
13 # Depth Anything V2 Base – Transformers Version
14
15 Depth Anything V2 is trained from 595K synthetic labeled images and 62M+ real unlabeled images, providing the most capable monocular depth estimation (MDE) model with the following features:
16 - more fine-grained details than Depth Anything V1
17 - more robust than Depth Anything V1 and SD-based models (e.g., Marigold, Geowizard)
18 - more efficient (10x faster) and more lightweight than SD-based models
19 - impressive fine-tuned performance with our pre-trained models
20
21 This model checkpoint is compatible with the transformers library.
22
23 Depth Anything V2 was introduced in [the paper of the same name](https://arxiv.org/abs/2406.09414) by Lihe Yang et al. It uses the same architecture as the original Depth Anything release, but uses synthetic data and a larger capacity teacher model to achieve much finer and robust depth predictions. The original Depth Anything model was introduced in the paper [Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data](https://arxiv.org/abs/2401.10891) by Lihe Yang et al., and was first released in [this repository](https://github.com/LiheYoung/Depth-Anything).
24
25 [Online demo](https://huggingface.co/spaces/depth-anything/Depth-Anything-V2).
26
27 ## Model description
28
29 Depth Anything V2 leverages the [DPT](https://huggingface.co/docs/transformers/model_doc/dpt) architecture with a [DINOv2](https://huggingface.co/docs/transformers/model_doc/dinov2) backbone.
30
31 The model is trained on ~600K synthetic labeled images and ~62 million real unlabeled images, obtaining state-of-the-art results for both relative and absolute depth estimation.
32
33 <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/depth_anything_overview.jpg"
34 alt="drawing" width="600"/>
35
36 <small> Depth Anything overview. Taken from the <a href="https://arxiv.org/abs/2401.10891">original paper</a>.</small>
37
38 ## Intended uses & limitations
39
40 You can use the raw model for tasks like zero-shot depth estimation. See the [model hub](https://huggingface.co/models?search=depth-anything) to look for
41 other versions on a task that interests you.
42
43 ### How to use
44
45 Here is how to use this model to perform zero-shot depth estimation:
46
47 ```python
48 from transformers import pipeline
49 from PIL import Image
50 import requests
51
52 # load pipe
53 pipe = pipeline(task="depth-estimation", model="depth-anything/Depth-Anything-V2-Large-hf")
54
55 # load image
56 url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
57 image = Image.open(requests.get(url, stream=True).raw)
58
59 # inference
60 depth = pipe(image)["depth"]
61 ```
62
63 Alternatively, you can use the model and processor classes:
64
65 ```python
66 from transformers import AutoImageProcessor, AutoModelForDepthEstimation
67 import torch
68 import numpy as np
69 from PIL import Image
70 import requests
71
72 url = "http://images.cocodataset.org/val2017/000000039769.jpg"
73 image = Image.open(requests.get(url, stream=True).raw)
74
75 image_processor = AutoImageProcessor.from_pretrained("depth-anything/Depth-Anything-V2-Large-hf")
76 model = AutoModelForDepthEstimation.from_pretrained("depth-anything/Depth-Anything-V2-Large-hf")
77
78 # prepare image for the model
79 inputs = image_processor(images=image, return_tensors="pt")
80
81 with torch.no_grad():
82 outputs = model(**inputs)
83 predicted_depth = outputs.predicted_depth
84
85 # interpolate to original size
86 prediction = torch.nn.functional.interpolate(
87 predicted_depth.unsqueeze(1),
88 size=image.size[::-1],
89 mode="bicubic",
90 align_corners=False,
91 )
92 ```
93
94 For more code examples, please refer to the [documentation](https://huggingface.co/transformers/main/model_doc/depth_anything.html#).
95
96
97 ### Citation
98
99 ```bibtex
100 @misc{yang2024depth,
101 title={Depth Anything V2},
102 author={Lihe Yang and Bingyi Kang and Zilong Huang and Zhen Zhao and Xiaogang Xu and Jiashi Feng and Hengshuang Zhao},
103 year={2024},
104 eprint={2406.09414},
105 archivePrefix={arXiv},
106 primaryClass={id='cs.CV' full_name='Computer Vision and Pattern Recognition' is_active=True alt_name=None in_archive='cs' is_general=False description='Covers image processing, computer vision, pattern recognition, and scene understanding. Roughly includes material in ACM Subject Classes I.2.10, I.4, and I.5.'}
107 }
108 ```
109