File size: 5,697 Bytes
96e9536
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
<!--Copyright 2023 The HuggingFace Team. All rights reserved.

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.

โš ๏ธ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.

-->

# ๋‹จ์ผ ์˜์ƒ ๊ธฐ๋ฐ˜ ๊นŠ์ด ์ถ”์ •[[depth-estimation-pipeline]]

๋‹จ์ผ ์˜์ƒ ๊ธฐ๋ฐ˜ ๊นŠ์ด ์ถ”์ •์€ ํ•œ ์žฅ๋ฉด์˜ ๋‹จ์ผ ์ด๋ฏธ์ง€์—์„œ ์žฅ๋ฉด์˜ ๊นŠ์ด ์ •๋ณด๋ฅผ ์˜ˆ์ธกํ•˜๋Š” ์ปดํ“จํ„ฐ ๋น„์ „ ์ž‘์—…์ž…๋‹ˆ๋‹ค.
์ฆ‰, ๋‹จ์ผ ์นด๋ฉ”๋ผ ์‹œ์ ์˜ ์žฅ๋ฉด์— ์žˆ๋Š” ๋ฌผ์ฒด์˜ ๊ฑฐ๋ฆฌ๋ฅผ ์˜ˆ์ธกํ•˜๋Š” ๊ณผ์ •์ž…๋‹ˆ๋‹ค.

๋‹จ์ผ ์˜์ƒ ๊ธฐ๋ฐ˜ ๊นŠ์ด ์ถ”์ •์€ 3D ์žฌ๊ตฌ์„ฑ, ์ฆ๊ฐ• ํ˜„์‹ค, ์ž์œจ ์ฃผํ–‰, ๋กœ๋ด‡ ๊ณตํ•™ ๋“ฑ ๋‹ค์–‘ํ•œ ๋ถ„์•ผ์—์„œ ์‘์šฉ๋ฉ๋‹ˆ๋‹ค. 
์กฐ๋ช… ์กฐ๊ฑด, ๊ฐ€๋ ค์ง, ํ…์Šค์ฒ˜์™€ ๊ฐ™์€ ์š”์†Œ์˜ ์˜ํ–ฅ์„ ๋ฐ›์„ ์ˆ˜ ์žˆ๋Š” ์žฅ๋ฉด ๋‚ด ๋ฌผ์ฒด์™€ ํ•ด๋‹น ๊นŠ์ด ์ •๋ณด ๊ฐ„์˜ ๋ณต์žกํ•œ ๊ด€๊ณ„๋ฅผ ๋ชจ๋ธ์ด ์ดํ•ดํ•ด์•ผ ํ•˜๋ฏ€๋กœ ๊นŒ๋‹ค๋กœ์šด ์ž‘์—…์ž…๋‹ˆ๋‹ค.


<Tip>
์ด ํŠœํ† ๋ฆฌ์–ผ์—์„œ ๋‹ค๋ฃจ๋Š” ์ž‘์—…์€ ๋‹ค์Œ ๋ชจ๋ธ ์•„ํ‚คํ…์ฒ˜์—์„œ ์ง€์›๋ฉ๋‹ˆ๋‹ค:

<!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->

[DPT](../model_doc/dpt), [GLPN](../model_doc/glpn)

<!--End of the generated tip-->

</Tip>

์ด๋ฒˆ ๊ฐ€์ด๋“œ์—์„œ ๋ฐฐ์šธ ๋‚ด์šฉ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์Šต๋‹ˆ๋‹ค:

* ๊นŠ์ด ์ถ”์ • ํŒŒ์ดํ”„๋ผ์ธ ๋งŒ๋“ค๊ธฐ
* ์ง์ ‘ ๊นŠ์ด ์ถ”์ • ์ถ”๋ก ํ•˜๊ธฐ

์‹œ์ž‘ํ•˜๊ธฐ ์ „์—, ํ•„์š”ํ•œ ๋ชจ๋“  ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๊ฐ€ ์„ค์น˜๋˜์–ด ์žˆ๋Š”์ง€ ํ™•์ธํ•˜์„ธ์š”:

```bash
pip install -q transformers
```

## ๊นŠ์ด ์ถ”์ • ํŒŒ์ดํ”„๋ผ์ธ[[depth-estimation-inference-by-hand]]

๊นŠ์ด ์ถ”์ •์„ ์ถ”๋ก ํ•˜๋Š” ๊ฐ€์žฅ ๊ฐ„๋‹จํ•œ ๋ฐฉ๋ฒ•์€ ํ•ด๋‹น ๊ธฐ๋Šฅ์„ ์ œ๊ณตํ•˜๋Š” [`pipeline`]์„ ์‚ฌ์šฉํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค.
[Hugging Face Hub ์ฒดํฌํฌ์ธํŠธ](https://huggingface.co/models?pipeline_tag=depth-estimation&sort=downloads)์—์„œ ํŒŒ์ดํ”„๋ผ์ธ์„ ์ดˆ๊ธฐํ™”ํ•ฉ๋‹ˆ๋‹ค:

```py
>>> from transformers import pipeline

>>> checkpoint = "vinvino02/glpn-nyu"
>>> depth_estimator = pipeline("depth-estimation", model=checkpoint)
```


๋‹ค์Œ์œผ๋กœ, ๋ถ„์„ํ•  ์ด๋ฏธ์ง€๋ฅผ ํ•œ ์žฅ ์„ ํƒํ•˜์„ธ์š”:

```py
>>> from PIL import Image
>>> import requests

>>> url = "https://unsplash.com/photos/HwBAsSbPBDU/download?ixid=MnwxMjA3fDB8MXxzZWFyY2h8MzR8fGNhciUyMGluJTIwdGhlJTIwc3RyZWV0fGVufDB8MHx8fDE2Nzg5MDEwODg&force=true&w=640"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> image
```

<div class="flex justify-center">
     <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/depth-estimation-example.jpg" alt="Photo of a busy street"/>
</div>

์ด๋ฏธ์ง€๋ฅผ ํŒŒ์ดํ”„๋ผ์ธ์œผ๋กœ ์ „๋‹ฌํ•ฉ๋‹ˆ๋‹ค.

```py
>>> predictions = depth_estimator(image)
```

ํŒŒ์ดํ”„๋ผ์ธ์€ ๋‘ ๊ฐœ์˜ ํ•ญ๋ชฉ์„ ๊ฐ€์ง€๋Š” ๋”•์…”๋„ˆ๋ฆฌ๋ฅผ ๋ฐ˜ํ™˜ํ•ฉ๋‹ˆ๋‹ค.
์ฒซ ๋ฒˆ์งธ๋Š” `predicted_depth`๋กœ ๊ฐ ํ”ฝ์…€์˜ ๊นŠ์ด๋ฅผ ๋ฏธํ„ฐ๋กœ ํ‘œํ˜„ํ•œ ๊ฐ’์„ ๊ฐ€์ง€๋Š” ํ…์„œ์ž…๋‹ˆ๋‹ค.
๋‘ ๋ฒˆ์งธ๋Š” `depth`๋กœ ๊นŠ์ด ์ถ”์ • ๊ฒฐ๊ณผ๋ฅผ ์‹œ๊ฐํ™”ํ•˜๋Š” PIL ์ด๋ฏธ์ง€์ž…๋‹ˆ๋‹ค.

์ด์ œ ์‹œ๊ฐํ™”ํ•œ ๊ฒฐ๊ณผ๋ฅผ ์‚ดํŽด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค:

```py
>>> predictions["depth"]
```

<div class="flex justify-center">
     <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/depth-visualization.png" alt="Depth estimation visualization"/>
</div>

## ์ง์ ‘ ๊นŠ์ด ์ถ”์ • ์ถ”๋ก ํ•˜๊ธฐ[[depth-estimation-inference-by-hand]]

์ด์ œ ๊นŠ์ด ์ถ”์ • ํŒŒ์ดํ”„๋ผ์ธ ์‚ฌ์šฉ๋ฒ•์„ ์‚ดํŽด๋ณด์•˜์œผ๋‹ˆ ๋™์ผํ•œ ๊ฒฐ๊ณผ๋ฅผ ๋ณต์ œํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ์‚ดํŽด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.
[Hugging Face Hub ์ฒดํฌํฌ์ธํŠธ](https://huggingface.co/models?pipeline_tag=depth-estimation&sort=downloads)์—์„œ ๋ชจ๋ธ๊ณผ ๊ด€๋ จ ํ”„๋กœ์„ธ์„œ๋ฅผ ๊ฐ€์ ธ์˜ค๋Š” ๊ฒƒ๋ถ€ํ„ฐ ์‹œ์ž‘ํ•ฉ๋‹ˆ๋‹ค.
์—ฌ๊ธฐ์„œ ์ด์ „์— ์‚ฌ์šฉํ•œ ์ฒดํฌํฌ์ธํŠธ์™€ ๋™์ผํ•œ ๊ฒƒ์„ ์‚ฌ์šฉํ•ฉ๋‹ˆ๋‹ค:

```py
>>> from transformers import AutoImageProcessor, AutoModelForDepthEstimation

>>> checkpoint = "vinvino02/glpn-nyu"

>>> image_processor = AutoImageProcessor.from_pretrained(checkpoint)
>>> model = AutoModelForDepthEstimation.from_pretrained(checkpoint)
```

ํ•„์š”ํ•œ ์ด๋ฏธ์ง€ ๋ณ€ํ™˜์„ ์ฒ˜๋ฆฌํ•˜๋Š” `image_processor`๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ๋ชจ๋ธ์— ๋Œ€ํ•œ ์ด๋ฏธ์ง€ ์ž…๋ ฅ์„ ์ค€๋น„ํ•ฉ๋‹ˆ๋‹ค.
`image_processor`๋Š” ํฌ๊ธฐ ์กฐ์ • ๋ฐ ์ •๊ทœํ™” ๋“ฑ ํ•„์š”ํ•œ ์ด๋ฏธ์ง€ ๋ณ€ํ™˜์„ ์ฒ˜๋ฆฌํ•ฉ๋‹ˆ๋‹ค:

```py
>>> pixel_values = image_processor(image, return_tensors="pt").pixel_values
```

์ค€๋น„ํ•œ ์ž…๋ ฅ์„ ๋ชจ๋ธ๋กœ ์ „๋‹ฌํ•ฉ๋‹ˆ๋‹ค:

```py
>>> import torch

>>> with torch.no_grad():
...     outputs = model(pixel_values)
...     predicted_depth = outputs.predicted_depth
```

๊ฒฐ๊ณผ๋ฅผ ์‹œ๊ฐํ™”ํ•ฉ๋‹ˆ๋‹ค:

```py
>>> import numpy as np

>>> # ์›๋ณธ ์‚ฌ์ด์ฆˆ๋กœ ๋ณต์›
>>> prediction = torch.nn.functional.interpolate(
...     predicted_depth.unsqueeze(1),
...     size=image.size[::-1],
...     mode="bicubic",
...     align_corners=False,
... ).squeeze()
>>> output = prediction.numpy()

>>> formatted = (output * 255 / np.max(output)).astype("uint8")
>>> depth = Image.fromarray(formatted)
>>> depth
```

<div class="flex justify-center">
     <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/depth-visualization.png" alt="Depth estimation visualization"/>
</div>