Skip to content

Commit 52cb0ea

Browse files
BreakthroughBreakthrough
committed
[docs] Update download page for 4.12.0-dev1 and convert docs to HTML
1 parent a1a2e91 commit 52cb0ea

File tree

8 files changed

+431
-199
lines changed

8 files changed

+431
-199
lines changed

‎.github/workflows/publish-docs.yml‎

Lines changed: 10 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,9 @@ on:
55
branches:
66
- 4.x-cuda
77
paths:
8-
- 'docs/**'
8+
- 'docs/**.html'
9+
- 'docs/**.css'
10+
- 'docs/**.js'
911
workflow_dispatch:
1012

1113
jobs:
@@ -26,6 +28,10 @@ jobs:
2628
git checkout gh-pages
2729
mv -f docs-tmp/* .
2830
git add *
29-
git commit -m "[docs] @${{ github.triggering_actor }}: Update Documentation" \
30-
-m "Source: ${{ github.ref_name }} (${{ github.sha }})"
31-
git push
31+
if git diff --cached --quiet; then
32+
echo "No changes to deploy."
33+
else
34+
git commit -m "[docs] @${{ github.triggering_actor }}: Update Documentation" \
35+
-m "Source: ${{ github.ref_name }} (${{ github.sha }})"
36+
git push
37+
fi

‎README.md‎

Lines changed: 2 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -1,10 +1,10 @@
11
## `opencv-python-cuda`
22

3-
Download: [Latest Release (4.12.0-dev0)](https://github.com/Breakthrough/opencv-python-cuda/releases/tag/4.12.0-dev0)
3+
Download: [Latest Release (4.12.0-dev1)](https://github.com/Breakthrough/opencv-python-cuda/releases/tag/4.12.0-dev1)
44

55
## What is `opencv-python-cuda`?
66

7-
Pre-built NVIDIA® CUDA™ enabled OpenCV packages for Python that come with all batteries included. This is a fork of [the official opencv-python project](https://github.com/opencv/opencv-python). Right now packages are only produced for Windows x64, and devices must be Turing class (GeForce 2000 series) or newer. Once installed via `pip` (or another Python package manager like `uv`), the following should *just work*:
7+
Pre-built NVIDIA® CUDA™ enabled OpenCV packages for Python that come with all batteries included. This is a fork of [the official opencv-python project](https://github.com/opencv/opencv-python). Right now packages are only produced for Windows x64, and devices must be Maxwell class (GeForce 900 series) or newer. Once installed via `pip` (or another Python package manager like `uv`), the following should *just work*:
88

99
```
1010
import cv2
@@ -52,15 +52,3 @@ All wheels are distributed with [FFmpeg](http://ffmpeg.org) licensed under the [
5252
Non-headless Linux wheels ship with [Qt 5](http://doc.qt.io/qt-5/lgpl.html) licensed under the [LGPLv3](http://www.gnu.org/licenses/lgpl-3.0.html).
5353

5454
The packages include also other binaries. Full list of licenses can be found from [LICENSE-3RD-PARTY.txt](https://github.com/opencv/opencv-python/blob/master/LICENSE-3RD-PARTY.txt).
55-
56-
### Supported Python versions
57-
58-
Python 3.x compatible pre-built wheels are provided for the officially supported Python versions (not in EOL):
59-
60-
- 3.7
61-
- 3.8
62-
- 3.9
63-
- 3.10
64-
- 3.11
65-
- 3.12
66-
- 3.13

‎docs/cuda-compatibility.html‎

Lines changed: 176 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,176 @@
1+
<!DOCTYPE html>
2+
<html lang="en">
3+
<head>
4+
<meta charset="UTF-8">
5+
<meta name="viewport" content="width=device-width, initial-scale=1.0">
6+
<title>CUDA Compatibility Reference | opencv-python-cuda</title>
7+
<link rel="stylesheet" href="style.css">
8+
<link rel="preconnect" href="https://fonts.googleapis.com">
9+
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
10+
<link href="https://fonts.googleapis.com/css2?family=Noto+Sans:wght@300;400;500;700&display=swap" rel="stylesheet">
11+
</head>
12+
<body>
13+
<header>
14+
<div class="container">
15+
<h1><a href="index.html">opencv-python-cuda</a></h1>
16+
<p>CUDA Compatibility Reference</p>
17+
</div>
18+
</header>
19+
20+
<main>
21+
<div class="container">
22+
<section class="content-section">
23+
<h2>CUDA Toolkit vs Compute Capability Support</h2>
24+
<table>
25+
<thead>
26+
<tr>
27+
<th>CUDA Version</th>
28+
<th>Min Compute</th>
29+
<th>Max Compute</th>
30+
<th>Notes</th>
31+
</tr>
32+
</thead>
33+
<tbody>
34+
<tr>
35+
<td>13.x</td>
36+
<td>7.5</td>
37+
<td>12.0+</td>
38+
<td>Drops Maxwell, Pascal, Volta</td>
39+
</tr>
40+
<tr>
41+
<td>12.x</td>
42+
<td>5.0</td>
43+
<td>10.0+</td>
44+
<td>Current LTS, drops Kepler</td>
45+
</tr>
46+
<tr>
47+
<td>11.x</td>
48+
<td>3.5</td>
49+
<td>8.6+</td>
50+
<td>Legacy, includes Kepler</td>
51+
</tr>
52+
</tbody>
53+
</table>
54+
</section>
55+
56+
<section class="content-section">
57+
<h2>GPU Architectures</h2>
58+
<table>
59+
<thead>
60+
<tr>
61+
<th>Compute Capability</th>
62+
<th>Architecture</th>
63+
<th>Example GPUs</th>
64+
<th>Released</th>
65+
</tr>
66+
</thead>
67+
<tbody>
68+
<tr>
69+
<td>10.0</td>
70+
<td>Blackwell</td>
71+
<td>RTX 50 series</td>
72+
<td>2025</td>
73+
</tr>
74+
<tr>
75+
<td>9.0</td>
76+
<td>Hopper</td>
77+
<td>H100, H200</td>
78+
<td>2022</td>
79+
</tr>
80+
<tr>
81+
<td>8.9</td>
82+
<td>Ada Lovelace</td>
83+
<td>RTX 40 series</td>
84+
<td>2022</td>
85+
</tr>
86+
<tr>
87+
<td>8.6</td>
88+
<td>Ampere (consumer)</td>
89+
<td>RTX 30 series</td>
90+
<td>2020</td>
91+
</tr>
92+
<tr>
93+
<td>8.0</td>
94+
<td>Ampere (datacenter)</td>
95+
<td>A100, A30</td>
96+
<td>2020</td>
97+
</tr>
98+
<tr>
99+
<td>7.5</td>
100+
<td>Turing</td>
101+
<td>RTX 20 series, GTX 16 series</td>
102+
<td>2018</td>
103+
</tr>
104+
<tr>
105+
<td>7.0</td>
106+
<td>Volta</td>
107+
<td>V100, Titan V</td>
108+
<td>2017</td>
109+
</tr>
110+
<tr>
111+
<td>6.1</td>
112+
<td>Pascal (consumer)</td>
113+
<td>GTX 10 series</td>
114+
<td>2016</td>
115+
</tr>
116+
<tr>
117+
<td>6.0</td>
118+
<td>Pascal (datacenter)</td>
119+
<td>P100</td>
120+
<td>2016</td>
121+
</tr>
122+
<tr>
123+
<td>5.2</td>
124+
<td>Maxwell v2</td>
125+
<td>GTX 900 series, Titan X</td>
126+
<td>2014</td>
127+
</tr>
128+
<tr>
129+
<td>5.0</td>
130+
<td>Maxwell v1</td>
131+
<td>GTX 750 Ti</td>
132+
<td>2014</td>
133+
</tr>
134+
<tr>
135+
<td>3.5</td>
136+
<td>Kepler</td>
137+
<td>GTX 600/700 series</td>
138+
<td>2012</td>
139+
</tr>
140+
</tbody>
141+
</table>
142+
</section>
143+
144+
<section class="content-section">
145+
<h2>SASS vs PTX</h2>
146+
<ul>
147+
<li><strong>SASS</strong> (Shader Assembly): Native GPU code, runs directly on matching architecture</li>
148+
<li><strong>PTX</strong> (Parallel Thread Execution): Intermediate code, JIT-compiled at runtime</li>
149+
</ul>
150+
<p>
151+
PTX is forward-compatible only (e.g. PTX 10.0 works on GPUs &ge;10.0, not older).
152+
SASS is required for older GPUs that are lower than the PTX version.
153+
A GPU needs either matching SASS or older PTX which it can JIT compile.
154+
</p>
155+
</section>
156+
157+
<section class="content-section">
158+
<h2>References</h2>
159+
<ul>
160+
<li><a href="https://developer.nvidia.com/cuda-toolkit-archive" target="_blank">CUDA Toolkit Archive</a></li>
161+
<li><a href="https://developer.nvidia.com/cuda-gpus" target="_blank">CUDA GPUs - Compute Capability</a></li>
162+
<li><a href="https://developer.nvidia.com/cudnn-archive" target="_blank">CuDNN Archive</a></li>
163+
<li><a href="https://developer.nvidia.com/video-codec-sdk" target="_blank">Video Codec SDK</a></li>
164+
</ul>
165+
</section>
166+
</div>
167+
</main>
168+
169+
<footer>
170+
<div class="container">
171+
<p>&copy; 2026 Breakthrough. All rights reserved. Based on the official <a href="https://github.com/opencv/opencv-python">opencv-python</a> project.</p>
172+
</div>
173+
</footer>
174+
175+
</body>
176+
</html>

‎docs/cuda-compatibility.md‎

Lines changed: 0 additions & 42 deletions
This file was deleted.

‎docs/index.html‎

Lines changed: 16 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -13,23 +13,19 @@
1313
<header>
1414
<div class="container">
1515
<h1>opencv-python-cuda</h1>
16-
<p>Pre-built NVIDIA® CUDA™ enabled OpenCV packages for Python.</p>
16+
<p>Standalone NVIDIA® CUDA™ enabled OpenCV packages for Python.</p>
1717
</div>
1818
</header>
1919

2020
<main>
2121
<div class="container">
2222
<section class="download-section">
2323
<h2>Download</h2>
24-
<p class="version">Latest Release: 4.12.0-dev0</p>
24+
<p class="version">Latest Release: 4.12.0-dev1</p>
2525
<div class="platform-selection">
2626
<div class="platform">
2727
<h3>Windows</h3>
28-
<a href="https://github.com/Breakthrough/opencv-python-cuda/releases/tag/4.12.0-dev0" class="download-button">x86_64</a>
29-
</div>
30-
<div class="platform">
31-
<h3>Linux</h3>
32-
<a href="#" class="download-button disabled">x86_64 (Coming Soon)</a>
28+
<a href="https://github.com/Breakthrough/opencv-python-cuda/releases/tag/4.12.0-dev1" class="download-button">x86_64</a>
3329
</div>
3430
</div>
3531
<div class="eula">
@@ -44,18 +40,30 @@ <h2>About opencv-python-cuda</h2>
4440
It is a fork of the official <a href="https://github.com/opencv/opencv-python">opencv-python</a> project.
4541
Currently, packages are only produced for Windows x64, and devices must be Maxwell class (GeForce 900 series) or newer.
4642
</p>
43+
<p>
44+
The wheel is fully standalone &mdash; no CUDA Toolkit, cuDNN, or any other NVIDIA software needs to be installed on your system.
45+
All required runtime libraries are bundled directly in the package. Just install the <code>.whl</code> file and you're ready to go.
46+
</p>
4747
<p>
4848
Once installed via <code>pip</code> (or another Python package manager like <code>uv</code>), the following should just work:
4949
</p>
5050
<pre><code>import cv2
5151
print(cv2.cuda.getCudaEnabledDeviceCount())</code></pre>
5252
</section>
53+
54+
<section class="about-section">
55+
<h2>Documentation</h2>
56+
<ul class="doc-links">
57+
<li><a href="cuda-compatibility.html">CUDA Compatibility Reference</a> &mdash; GPU architectures, compute capabilities, and CUDA toolkit support</li>
58+
<li><a href="workflow.html">Workflow Guide</a> &mdash; Repository setup, triggering builds, and wheel compression</li>
59+
</ul>
60+
</section>
5361
</div>
5462
</main>
5563

5664
<footer>
5765
<div class="container">
58-
<p>&copy; 2025 Breakthrough. All rights reserved. Based on the official <a href="https://github.com/opencv/opencv-python">opencv-python</a> project.</p>
66+
<p>&copy; 2026 Breakthrough. All rights reserved. Based on the official <a href="https://github.com/opencv/opencv-python">opencv-python</a> project.</p>
5967
</div>
6068
</footer>
6169

0 commit comments

Comments
 (0)