# c-yn/adair architecture

> Architecture diagram of the GitHub repository c-yn/adair, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-02). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/c-yn/adair
- Repository: https://github.com/c-yn/adair

## Overview

AdaIR is an all-in-one image-restoration research codebase: it trains one adaptive model across denoising, deraining, dehazing, deblurring, and enhancement, and provides task-specific and multi-degradation evaluation. The main documented workflow is configured training data → degraded/clean patches → model restoration and loss optimization; evaluation uses a selected task mode and checkpoint. The model’s frequency-mining, modulation, and progressive-restoration stages are documented in the README. The evaluation implementation and some model internals were not sampled, so those links reflect README statements rather than inspected calls.

Main entry points: [`train.py`](https://github.com/c-yn/adair/blob/main/train.py) (Training loop), [`utils/dataset_utils.py`](https://github.com/c-yn/adair/blob/main/utils/dataset_utils.py) (Training dataset), [`test.py`](https://github.com/c-yn/adair/blob/main/test.py) (Task evaluation), [`options.py`](https://github.com/c-yn/adair/blob/main/options.py) (Run options)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_configuration["Configuration"]
  node_options["Run options<br/>[options.py]"]
end

subgraph group_data["Data preparation"]
  node_train_data["Training dataset<br/>[dataset_utils.py]"]
  node_degradation["Degradation synthesis"]
  node_image_utils["Image transforms<br/>[image_utils.py]"]
end

subgraph group_model["Restoration model"]
  node_encoder["Feature encoder<br/>[model.py]"]
  node_frequency["Frequency mining<br/>[model.py]"]
  node_modulation["Bidirectional modulation<br/>[model.py]"]
  node_decoder["Progressive restoration<br/>[model.py]"]
end

subgraph group_runtime["Training and evaluation"]
  node_trainer["Training loop<br/>[train.py]"]
  node_schedule["Learning-rate schedule<br/>[schedulers.py]"]
  node_evaluator["Task evaluation<br/>[test.py]"]
end

node_user(("Researcher"))
node_training_images[("Training images")]
node_checkpoint[("Model checkpoint")]
node_restored["Restored image"]
node_test_data[("Evaluation images")]

node_user -->|"sets options"| node_options
node_user -->|"prepares data"| node_training_images
node_options -->|"configures"| node_trainer
node_trainer -->|"constructs dataset"| node_train_data
node_train_data -.->|"reads samples"| node_training_images
node_train_data -->|"uses"| node_degradation
node_train_data -->|"uses transforms"| node_image_utils
node_trainer -->|"invokes model"| node_encoder
node_trainer -->|"configures schedule"| node_schedule
node_encoder -.->|"extracts features"| node_frequency
node_frequency -.->|"feeds modulation"| node_modulation
node_modulation -.->|"guides restoration"| node_decoder
node_decoder -.->|"produces"| node_restored
node_trainer -->|"saves checkpoints"| node_checkpoint
node_user -->|"prepares inputs"| node_test_data
node_user -->|"selects task mode"| node_evaluator
node_evaluator -->|"loads checkpoint"| node_checkpoint
node_evaluator -->|"evaluates images"| node_test_data
node_evaluator -.->|"restores images"| node_restored

click node_options "https://github.com/c-yn/adair/blob/main/options.py"
click node_train_data "https://github.com/c-yn/adair/blob/main/utils/dataset_utils.py"
click node_degradation "https://github.com/c-yn/adair/blob/main/utils/degradation_utils.py"
click node_image_utils "https://github.com/c-yn/adair/blob/main/utils/image_utils.py"
click node_trainer "https://github.com/c-yn/adair/blob/main/train.py"
click node_schedule "https://github.com/c-yn/adair/blob/main/utils/schedulers.py"
click node_encoder "https://github.com/c-yn/adair/blob/main/net/model.py"
click node_frequency "https://github.com/c-yn/adair/blob/main/net/model.py"
click node_modulation "https://github.com/c-yn/adair/blob/main/net/model.py"
click node_decoder "https://github.com/c-yn/adair/blob/main/net/model.py"
click node_evaluator "https://github.com/c-yn/adair/blob/main/test.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_options toneBlue
class node_train_data,node_degradation,node_image_utils,node_training_images,node_checkpoint,node_test_data toneAmber
class node_encoder,node_frequency,node_modulation,node_decoder toneMint
class node_trainer,node_schedule,node_evaluator toneRose
class node_user,node_restored toneIndigo
```

## Components

### Configuration

- [Run options](https://github.com/c-yn/adair/blob/main/options.py): `options.py`

### Data preparation

- [Training dataset](https://github.com/c-yn/adair/blob/main/utils/dataset_utils.py): `utils/dataset_utils.py`
- [Degradation synthesis](https://github.com/c-yn/adair/blob/main/utils/degradation_utils.py): `utils/degradation_utils.py`
- [Image transforms](https://github.com/c-yn/adair/blob/main/utils/image_utils.py): `utils/image_utils.py`

### Restoration model

- [Feature encoder](https://github.com/c-yn/adair/blob/main/net/model.py): `net/model.py`
- [Frequency mining](https://github.com/c-yn/adair/blob/main/net/model.py): `net/model.py`
- [Bidirectional modulation](https://github.com/c-yn/adair/blob/main/net/model.py): `net/model.py`
- [Progressive restoration](https://github.com/c-yn/adair/blob/main/net/model.py): `net/model.py`

### Training and evaluation

- [Training loop](https://github.com/c-yn/adair/blob/main/train.py): `train.py`
- [Learning-rate schedule](https://github.com/c-yn/adair/blob/main/utils/schedulers.py): `utils/schedulers.py`
- [Task evaluation](https://github.com/c-yn/adair/blob/main/test.py): `test.py`

### Other components

- **Researcher**
- **Training images**
- **Model checkpoint**
- **Restored image**
- **Evaluation images**

## Connections

- Researcher → Run options: sets options (evidence: `options.py`)
- Researcher → Training images: prepares data (evidence: `README.md`)
- Run options → Training loop: configures (evidence: `train.py`)
- Training loop → Training dataset: constructs dataset (evidence: `train.py`)
- Training dataset → Training images: reads samples
- Training dataset → Degradation synthesis: uses (evidence: `utils/dataset_utils.py`)
- Training dataset → Image transforms: uses transforms (evidence: `utils/dataset_utils.py`)
- Training loop → Feature encoder: invokes model (evidence: `train.py`)
- Training loop → Learning-rate schedule: configures schedule (evidence: `train.py`)
- Feature encoder → Frequency mining: extracts features (evidence: `README.md`)
- Frequency mining → Bidirectional modulation: feeds modulation (evidence: `README.md`)
- Bidirectional modulation → Progressive restoration: guides restoration (evidence: `README.md`)
- Progressive restoration → Restored image: produces (evidence: `README.md`)
- Training loop → Model checkpoint: saves checkpoints (evidence: `train.py`)
- Researcher → Evaluation images: prepares inputs (evidence: `README.md`)
- Researcher → Task evaluation: selects task mode (evidence: `README.md`)
- Task evaluation → Model checkpoint: loads checkpoint (evidence: `README.md`)
- Task evaluation → Evaluation images: evaluates images (evidence: `README.md`)
- Task evaluation → Restored image: restores images
