So here are very slowly processed close-ups of NGC7331, with and without Graxpert denoising. Small but noticeable improvement with denoising I think
🔭
So here are very slowly processed close-ups of NGC7331, with and without Graxpert denoising. Small but noticeable improvement with denoising I think
🔭
💥 We've added a lightweight multi-branch and multi-scale feature extraction network for superior visual quality.
Our recently expanded blog has all the details 👉 gpuopen.com/learn/neural...
💥 We've added a lightweight multi-branch and multi-scale feature extraction network for superior visual quality.
Our recently expanded blog has all the details 👉 gpuopen.com/learn/neural...
Combining the results of multiple denoising models by taking the minimum value at each pixel helps.
github.com/olibclarke/c...
Combining the results of multiple denoising models by taking the minimum value at each pixel helps.
github.com/olibclarke/c...
this is after deinterlacing and denoising so not the raw image but like holy shit
this is after deinterlacing and denoising so not the raw image but like holy shit
- Music
- Denoising
- Intro rewriting (due to developing current events)
- Minor re-records
- One reshoot (ffs)
- Fixing early citation graphics
- Creating a montage
it's 4 hours long & doesn't have credits yet btw.
- Music
- Denoising
- Intro rewriting (due to developing current events)
- Minor re-records
- One reshoot (ffs)
- Fixing early citation graphics
- Creating a montage
it's 4 hours long & doesn't have credits yet btw.
gpuopen.com/learn/neural...
gpuopen.com/learn/neural...
Parallel mirrors. Tracing 8 paths/pixel per frame + DNN denoising. No temporal accumulation.
Parallel mirrors. Tracing 8 paths/pixel per frame + DNN denoising. No temporal accumulation.