| <!DOCTYPE html> |
| <html> |
| <head> |
| <meta charset="utf-8"> |
| <title>PySIFT: GPU-Resident Deterministic SIFT</title> |
| <style> |
| body { font-family: system-ui, sans-serif; max-width: 800px; margin: 40px auto; padding: 0 20px; color: #333; |
| } |
| h1 { color: #1a1a1a; } |
| .badge { display: inline-block; margin-right: 8px; } |
| .links { margin: 24px 0; } |
| .links a { display: inline-block; padding: 10px 20px; margin: 4px 8px 4px 0; background: #2ecc71; color: |
| white; text-decoration: none; border-radius: 6px; font-weight: bold; } |
| .links a.paper { background: #b31b1b; } |
| .links a.pypi { background: #3775a9; } |
| table { border-collapse: collapse; margin: 20px 0; } |
| th, td { border: 1px solid #ddd; padding: 8px 14px; text-align: left; } |
| th { background: #f5f5f5; } |
| code { background: #f0f0f0; padding: 2px 6px; border-radius: 3px; } |
| </style> |
| </head> |
| <body> |
| <h1>PySIFT</h1> |
| <p><strong>GPU-Resident Deterministic SIFT for Deep Learning Vision Pipelines</strong></p> |
|
|
| <div class="links"> |
| |
| <a class="paper" href="https://arxiv.org/abs/2605.17869" target="_blank" rel="noopener">arXiv Paper</a> |
| <a href="https://github.com/SivaIITM/PySIFT" target="_blank" rel="noopener">GitHub Code</a> |
| <a class="pypi" href="https://pypi.org/project/staysift/" target="_blank" rel="noopener">pip install staysift</a> |
| <a href="https://www.kaggle.com/code/sivakumarksce24d040/pysift-tutorial" target="_blank" rel="noopener" style="background-color:#20BEFF;color:white;padding:10px 20px;text-decoration:none;border-radius:5px;font-weight:bold;">Kaggle Tutorial</a> |
| </div> |
|
|
| <p>A pure-Python, GPU-resident SIFT implementation that matches OpenCV SIFT accuracy while running <strong>26% |
| faster end-to-end</strong> with <strong>4x matching speedup</strong>. Zero-copy DLPack interop keeps tensors on the |
| GPU across the full pipeline.</p> |
|
|
| <table> |
| <tr><th>Benchmark</th><th>Metric</th><th>PySIFT vs OpenCV</th></tr> |
| <tr><td>HPatches</td><td>MMA@10</td><td>+2.2pp</td></tr> |
| <tr><td>IMC Phototourism</td><td>Inliers/pair</td><td>303 vs 205 (+47%)</td></tr> |
| <tr><td>MegaDepth-1500</td><td>AUC@10</td><td>+5.6pp</td></tr> |
| <tr><td>ROxford5K</td><td>mAP</td><td>+7.5pp</td></tr> |
| </table> |
|
|
| <h3>Quick Start</h3> |
| <pre><code>pip install staysift |
| from pysift import PySIFT |
| sift = PySIFT() |
| keypoints, descriptors = sift.detectAndCompute(gray_image)</code></pre> |
| </body> |
| </html> |
|
|
