Another little area where we've been looking at the use of machine learning is for image denoising. What follows is still a work in progress, but overall things seem to work sufficiently well that some folks might be interested to give it a test drive.
A little bit of expectation management before you start downloading stuff:
- The model operates on full resolution, fully processed RGB images taken with Raspberry Pi cameras.
- The images should be captured with no hardware denoise or sharpening applied during the capture.
- It's slow to process an image (12 seconds per MP on a Pi 5), though you can run it on any computer you have.
- There are certainly some deficiencies in what it does sometimes!
Finally, the model and code are here.
Hopefully most of that is reasonably clear, but if you have any questions, please feel free to ask below. Thanks!
A little bit of expectation management before you start downloading stuff:
- The model operates on full resolution, fully processed RGB images taken with Raspberry Pi cameras.
- The images should be captured with no hardware denoise or sharpening applied during the capture.
- It's slow to process an image (12 seconds per MP on a Pi 5), though you can run it on any computer you have.
- There are certainly some deficiencies in what it does sometimes!
Finally, the model and code are here.
Hopefully most of that is reasonably clear, but if you have any questions, please feel free to ask below. Thanks!
Statistics: Posted by therealdavidp — Mon Apr 07, 2025 3:04 pm — Replies 0 — Views 23
