Downloads · 30 days
0
dlp-nppe3-denoise-sr
jagannath-r/dlp-nppe3-denoise-sr
dlp-nppe3-denoise-sr is a machine learning model from jagannath-r. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Residual CNN with PixelShuffle upsampling. Takes a 256x160 noisy low-light RGB image and outputs a 1024x640 denoised, super-resolved RGB image (4x upsampling in each dimension).
Access
Public
Updated Sep 5, 2026
Repo size
9.8 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pth6.1 MB · 100%
At a glance
- Access
- Public
- Created
- Sep 5, 2026
- Updated
- Sep 5, 2026
- SHA
- e6bfd772
Model card
From the Hugging Face model README
DLP NPPE-3: Low-Light Denoising + 4x Super-Resolution
Residual CNN with PixelShuffle upsampling. Takes a 256x160 noisy low-light RGB image and outputs a 1024x640 denoised, super-resolved RGB image (4x upsampling in each dimension).
- Architecture: 8 residual blocks, 64 feature channels, global bicubic-residual connection
- Loss: L1
- Training: 30 epochs, Adam (lr=1e-4), cosine annealing schedule, 64x64 LR patch crops
- Validation PSNR: 38.03 dB
- Created
- Sep 5, 2026
- Updated
- Sep 5, 2026