search

google-research / omniglue

Code release for CVPR'24 submission 'OmniGlue'
https://hwjiang1510.github.io/OmniGlue
Apache License 2.0
532 stars 45 forks source link

Add issue solutions and superpoint pytorch version that eat up excessive inference memory #23

Open Eunchan24 opened 3 months ago

Eunchan24 commented 3 months ago

Hello, thank you for your hard work! I suggest a minor code correction.

  1. Improved GPU selection Currently, many people lack GPU memory or have problems requiring too much GPU memory, so the model is not working. 😭 To resolve this issue, add the get_device_framework function to utils.py , which is not a perfect solution, but works more flexibly than before.

  2. Add Pytorch SuperPoint Add the Pytorch version of the SuperPoint and SuperPoint weight DownloadPath

  3. Changed the code style to PEP8 due to an indentation error in the omniglue_extract.py file. Let me know what you think 🙂