Low-Light Denoising + 4x Super-Resolution

This repository contains a custom PyTorch RRDBNet model fine-tuned for low-light image denoising and 4x super-resolution.

Model details

• Architecture: RRDBNet
• Input: RGB image
• Output: RGB image
• Upscaling factor: 4x
• Input range: [0, 1]
• RRDB blocks: 23
• Feature channels: 64
• Growth channels: 32

Files

• model.safetensors: fine-tuned model weights
• modeling_rrdbnet.py: model architecture
• config.json: model configuration

Loading the model

import torch
from safetensors.torch import load_file
from modeling_rrdbnet import RRDBNet

model = RRDBNet(
    num_in_ch=3,
    num_out_ch=3,
    num_feat=64,
    num_block=23,
    num_grow_ch=32,
)

model.load_state_dict(
    load_file("model.safetensors")
)

model.eval()

The model expects RGB images converted to float tensors in the [0, 1] range and outputs images at 4x the input resolution.