# megvii-research/nafnet architecture

> Architecture diagram of the GitHub repository megvii-research/nafnet, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-04). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/megvii-research/nafnet
- Repository: https://github.com/megvii-research/nafnet

## Overview

NAFNet is a PyTorch image-restoration toolbox for denoising, deblurring, and stereo super-resolution. The main user workflow is to provide degraded image(s), run an inference entry point, and save restored output. Training adds paired-image data preparation, model optimization, and validation metrics. The graph also captures reusable architecture, data, and runtime-support layers. Relationships are limited to those stated in the README or supported by the supplied source index; unsampled implementation details are not inferred.

Main entry points: [`basicsr/train.py`](https://github.com/megvii-research/nafnet/blob/main/basicsr/train.py) (Training loop), [`basicsr/models/image_restoration_model.py`](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/image_restoration_model.py) (Restoration task), [`basicsr/data`](https://github.com/megvii-research/nafnet/tree/main/basicsr/data) (Image and video datasets), [`basicsr/models/archs/NAFNet_arch.py`](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/NAFNet_arch.py) (NAFNet)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_entry["User workflows"]
  node_image_demo["Image inference<br/>[demo.py]"]
  node_stereo_demo["Stereo inference<br/>[demo_ssr.py]"]
  node_predict["Prediction entry<br/>[predict.py]"]
end

subgraph group_data["Data pipeline"]
  node_datasets["Image and video datasets"]
  node_data_util["Dataset utilities<br/>[data_util.py]"]
  node_transforms["Image transforms<br/>[transforms.py]"]
  node_prefetch["Data prefetch"]
end

subgraph group_model["Restoration models"]
  node_restoration["Restoration task"]
  node_nafnet["NAFNet<br/>[NAFNet_arch.py]"]
  node_baseline["Baseline network<br/>[Baseline_arch.py]"]
  node_nafssr["NAFSSR<br/>[NAFSSR_arch.py]"]
  node_arch_util["Architecture helpers<br/>[arch_util.py]"]
  node_local_arch["Local architecture<br/>[local_arch.py]"]
  node_losses["Training losses<br/>[losses.py]"]
end

subgraph group_runtime["Training and evaluation"]
  node_train["Training loop<br/>[train.py]"]
  node_base_model["Model base<br/>[base_model.py]"]
  node_metrics["Quality metrics"]
  node_options["Option parsing<br/>[options.py]"]
  node_file_client["Storage access<br/>[file_client.py]"]
  node_distributed["Distributed runtime<br/>[dist_util.py]"]
  node_logger["Logging<br/>[logger.py]"]
  node_matlab["Image operations"]
end

node_user(("Image user"))

node_user -->|"runs inference"| node_image_demo
node_user -->|"runs stereo"| node_stereo_demo
node_user -.->|"runs prediction"| node_predict
node_image_demo -->|"restores images"| node_restoration
node_stereo_demo -->|"runs stereo SR"| node_nafssr
node_predict -->|"creates model"| node_restoration
node_train -->|"parses options"| node_options
node_train -->|"loads data"| node_datasets
node_train -->|"uses prefetch"| node_prefetch
node_train -->|"creates model"| node_restoration
node_train -->|"initializes distributed"| node_distributed
node_train -.->|"sets up logging"| node_logger
node_train -.->|"evaluates output"| node_metrics
node_datasets -->|"uses utilities"| node_data_util
node_datasets -->|"uses transforms"| node_transforms
node_restoration -->|"inherits from"| node_base_model
node_restoration -->|"builds network"| node_nafnet
node_restoration -->|"configures losses"| node_losses
node_nafnet -->|"uses helpers"| node_arch_util
node_nafnet -->|"uses modules"| node_local_arch
node_baseline -->|"uses helpers"| node_arch_util
node_baseline -->|"uses modules"| node_local_arch
node_metrics -->|"uses image ops"| node_matlab
node_datasets -.->|"reads files"| node_file_client

click node_image_demo "https://github.com/megvii-research/nafnet/blob/main/basicsr/demo.py"
click node_stereo_demo "https://github.com/megvii-research/nafnet/blob/main/basicsr/demo_ssr.py"
click node_predict "https://github.com/megvii-research/nafnet/blob/main/predict.py"
click node_restoration "https://github.com/megvii-research/nafnet/blob/main/basicsr/models/image_restoration_model.py"
click node_nafnet "https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/NAFNet_arch.py"
click node_baseline "https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/Baseline_arch.py"
click node_nafssr "https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/NAFSSR_arch.py"
click node_arch_util "https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/arch_util.py"
click node_local_arch "https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/local_arch.py"
click node_losses "https://github.com/megvii-research/nafnet/blob/main/basicsr/models/losses/losses.py"
click node_datasets "https://github.com/megvii-research/nafnet/tree/main/basicsr/data"
click node_data_util "https://github.com/megvii-research/nafnet/blob/main/basicsr/data/data_util.py"
click node_transforms "https://github.com/megvii-research/nafnet/blob/main/basicsr/data/transforms.py"
click node_prefetch "https://github.com/megvii-research/nafnet/blob/main/basicsr/data/prefetch_dataloader.py"
click node_train "https://github.com/megvii-research/nafnet/blob/main/basicsr/train.py"
click node_base_model "https://github.com/megvii-research/nafnet/blob/main/basicsr/models/base_model.py"
click node_metrics "https://github.com/megvii-research/nafnet/tree/main/basicsr/metrics"
click node_options "https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/options.py"
click node_file_client "https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/file_client.py"
click node_distributed "https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/dist_util.py"
click node_logger "https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/logger.py"
click node_matlab "https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/matlab_functions.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_image_demo,node_stereo_demo,node_predict,node_user toneBlue
class node_datasets,node_data_util,node_transforms,node_prefetch toneAmber
class node_restoration,node_nafnet,node_baseline,node_nafssr,node_arch_util,node_local_arch,node_losses toneMint
class node_train,node_base_model,node_metrics,node_options,node_file_client,node_distributed,node_logger,node_matlab toneRose
```

## Components

### User workflows

- [Image inference](https://github.com/megvii-research/nafnet/blob/main/basicsr/demo.py): `basicsr/demo.py`
- [Stereo inference](https://github.com/megvii-research/nafnet/blob/main/basicsr/demo_ssr.py): `basicsr/demo_ssr.py`
- [Prediction entry](https://github.com/megvii-research/nafnet/blob/main/predict.py): `predict.py`

### Data pipeline

- [Image and video datasets](https://github.com/megvii-research/nafnet/tree/main/basicsr/data): `basicsr/data`
- [Dataset utilities](https://github.com/megvii-research/nafnet/blob/main/basicsr/data/data_util.py): `basicsr/data/data_util.py`
- [Image transforms](https://github.com/megvii-research/nafnet/blob/main/basicsr/data/transforms.py): `basicsr/data/transforms.py`
- [Data prefetch](https://github.com/megvii-research/nafnet/blob/main/basicsr/data/prefetch_dataloader.py): `basicsr/data/prefetch_dataloader.py`

### Restoration models

- [Restoration task](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/image_restoration_model.py): `basicsr/models/image_restoration_model.py`
- [NAFNet](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/NAFNet_arch.py): `basicsr/models/archs/NAFNet_arch.py`
- [Baseline network](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/Baseline_arch.py): `basicsr/models/archs/Baseline_arch.py`
- [NAFSSR](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/NAFSSR_arch.py): `basicsr/models/archs/NAFSSR_arch.py`
- [Architecture helpers](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/arch_util.py): `basicsr/models/archs/arch_util.py`
- [Local architecture](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/archs/local_arch.py): `basicsr/models/archs/local_arch.py`
- [Training losses](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/losses/losses.py): `basicsr/models/losses/losses.py`

### Training and evaluation

- [Training loop](https://github.com/megvii-research/nafnet/blob/main/basicsr/train.py): `basicsr/train.py`
- [Model base](https://github.com/megvii-research/nafnet/blob/main/basicsr/models/base_model.py): `basicsr/models/base_model.py`
- [Quality metrics](https://github.com/megvii-research/nafnet/tree/main/basicsr/metrics): `basicsr/metrics`
- [Option parsing](https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/options.py): `basicsr/utils/options.py`
- [Storage access](https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/file_client.py): `basicsr/utils/file_client.py`
- [Distributed runtime](https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/dist_util.py): `basicsr/utils/dist_util.py`
- [Logging](https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/logger.py): `basicsr/utils/logger.py`
- [Image operations](https://github.com/megvii-research/nafnet/blob/main/basicsr/utils/matlab_functions.py): `basicsr/utils/matlab_functions.py`

### Other components

- **Image user**

## Connections

- Image user → Image inference: runs inference
- Image user → Stereo inference: runs stereo
- Image user → Prediction entry: runs prediction
- Image inference → Restoration task: restores images
- Stereo inference → NAFSSR: runs stereo SR
- Prediction entry → Restoration task: creates model
- Training loop → Option parsing: parses options (evidence: `basicsr/train.py`)
- Training loop → Image and video datasets: loads data (evidence: `basicsr/train.py`)
- Training loop → Data prefetch: uses prefetch (evidence: `basicsr/train.py`)
- Training loop → Restoration task: creates model
- Training loop → Distributed runtime: initializes distributed (evidence: `basicsr/train.py`)
- Training loop → Logging: sets up logging
- Training loop → Quality metrics: evaluates output
- Image and video datasets → Dataset utilities: uses utilities (evidence: `basicsr/data/video_test_dataset.py`)
- Image and video datasets → Image transforms: uses transforms (evidence: `basicsr/data/data_util.py`)
- Restoration task → Model base: inherits from (evidence: `basicsr/models/image_restoration_model.py`)
- Restoration task → NAFNet: builds network
- Restoration task → Training losses: configures losses (evidence: `basicsr/models/image_restoration_model.py`)
- NAFNet → Architecture helpers: uses helpers (evidence: `basicsr/models/archs/NAFNet_arch.py`)
- NAFNet → Local architecture: uses modules (evidence: `basicsr/models/archs/NAFNet_arch.py`)
- Baseline network → Architecture helpers: uses helpers (evidence: `basicsr/models/archs/Baseline_arch.py`)
- Baseline network → Local architecture: uses modules (evidence: `basicsr/models/archs/Baseline_arch.py`)
- Quality metrics → Image operations: uses image ops (evidence: `basicsr/metrics/metric_util.py`)
- Image and video datasets → Storage access: reads files
