# csbdeep/csbdeep architecture

> Architecture diagram of the GitHub repository csbdeep/csbdeep, 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/csbdeep/csbdeep
- Repository: https://github.com/csbdeep/csbdeep

## Overview

CSBDeep is a Python toolbox for CARE deep-learning restoration of fluorescence microscopy images. Its main workflow prepares image pairs and patches, trains a configured CARE model, then predicts restored images; model variants support isotropic, projection, and upsampling use cases. The graph includes the model-building and prediction internals, pretrained-model support, and the command-line interface. Relationships are limited to those documented in the README or visible in the supplied excerpts and SOURCE INDEX; unsampled details are not inferred.

Main entry points: [`csbdeep/models/care_standard.py`](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_standard.py) (CARE training and prediction), [`csbdeep/scripts/care_predict.py`](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/scripts/care_predict.py) (Prediction CLI), [`csbdeep/internals/predict.py`](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/predict.py) (Tiled prediction), [`csbdeep/models/care_isotropic.py`](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_isotropic.py) (Isotropic CARE)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_data["Data preparation"]
  node_rawdata["Raw image pairs<br/>[rawdata.py]"]
  node_transforms["Image transforms<br/>[transform.py]"]
  node_patches["Patch generation<br/>[generate.py]"]
  node_prepare["Data preparation<br/>[prepare.py]"]
end

subgraph group_models["CARE models"]
  node_config["Model configuration<br/>[config.py]"]
  node_base["Model lifecycle<br/>[base_model.py]"]
  node_care["CARE training and prediction<br/>[care_standard.py]"]
  node_isotropic["Isotropic CARE<br/>[care_isotropic.py]"]
  node_projection["Projection CARE<br/>[care_projection.py]"]
  node_upsampling["Upsampling CARE<br/>[care_upsampling.py]"]
  node_pretrained["Pretrained models<br/>[pretrained.py]"]
end

subgraph group_engine["Model internals"]
  node_nets["Network builders<br/>[nets.py]"]
  node_blocks["Network blocks<br/>[blocks.py]"]
  node_training["Training engine<br/>[train.py]"]
  node_losses["Training losses<br/>[losses.py]"]
  node_prediction["Tiled prediction<br/>[predict.py]"]
  node_probability["Probabilistic output<br/>[probability.py]"]
end

subgraph group_interfaces["Input and interfaces"]
  node_io["Image file I/O<br/>[__init__.py]"]
  node_cli["Prediction CLI<br/>[care_predict.py]"]
  node_restored["Restored images<br/>[__init__.py]"]
end

node_user(("Microscopy user"))

node_user -.->|"provides images"| node_rawdata
node_rawdata -.->|"supplies image pairs"| node_patches
node_patches -->|"applies transforms"| node_transforms
node_patches -->|"saves patches"| node_io
node_user -.->|"configures model"| node_config
node_config -->|"configures lifecycle"| node_base
node_base -->|"uses pretrained support"| node_pretrained
node_care -->|"extends lifecycle"| node_base
node_care -->|"uses configuration"| node_config
node_care -->|"builds network"| node_nets
node_care -->|"trains with"| node_training
node_care -->|"predicts with"| node_prediction
node_care -->|"returns distributions"| node_probability
node_nets -->|"uses blocks"| node_blocks
node_training -->|"uses losses"| node_losses
node_isotropic -->|"extends CARE"| node_care
node_isotropic -->|"uses prediction"| node_prediction
node_isotropic -->|"uses distributions"| node_probability
node_projection -->|"extends CARE"| node_care
node_projection -->|"uses network builder"| node_nets
node_projection -->|"uses prediction"| node_prediction
node_upsampling -->|"extends CARE"| node_care
node_cli -->|"reads and writes images"| node_io
node_cli -.->|"writes results"| node_restored
node_cli -->|"invokes model"| node_care
node_cli -.->|"loads data"| node_prepare
node_prediction -.->|"produces predictions"| node_restored

click node_rawdata "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/data/rawdata.py"
click node_transforms "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/data/transform.py"
click node_patches "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/data/generate.py"
click node_prepare "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/data/prepare.py"
click node_io "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/io/__init__.py"
click node_config "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/config.py"
click node_base "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/base_model.py"
click node_care "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_standard.py"
click node_isotropic "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_isotropic.py"
click node_projection "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_projection.py"
click node_upsampling "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_upsampling.py"
click node_pretrained "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/pretrained.py"
click node_nets "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/nets.py"
click node_blocks "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/blocks.py"
click node_training "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/train.py"
click node_losses "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/losses.py"
click node_prediction "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/predict.py"
click node_probability "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/probability.py"
click node_cli "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/scripts/care_predict.py"
click node_restored "https://github.com/csbdeep/csbdeep/blob/main/csbdeep/io/__init__.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_rawdata,node_transforms,node_patches,node_prepare,node_user toneBlue
class node_config,node_base,node_care,node_isotropic,node_projection,node_upsampling,node_pretrained toneAmber
class node_nets,node_blocks,node_training,node_losses,node_prediction,node_probability toneMint
class node_io,node_cli,node_restored toneRose
```

## Components

### Data preparation

- [Raw image pairs](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/data/rawdata.py): `csbdeep/data/rawdata.py`
- [Image transforms](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/data/transform.py): `csbdeep/data/transform.py`
- [Patch generation](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/data/generate.py): `csbdeep/data/generate.py`
- [Data preparation](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/data/prepare.py): `csbdeep/data/prepare.py`

### CARE models

- [Model configuration](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/config.py): `csbdeep/models/config.py`
- [Model lifecycle](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/base_model.py): `csbdeep/models/base_model.py`
- [CARE training and prediction](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_standard.py): `csbdeep/models/care_standard.py`
- [Isotropic CARE](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_isotropic.py): `csbdeep/models/care_isotropic.py`
- [Projection CARE](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_projection.py): `csbdeep/models/care_projection.py`
- [Upsampling CARE](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/care_upsampling.py): `csbdeep/models/care_upsampling.py`
- [Pretrained models](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/models/pretrained.py): `csbdeep/models/pretrained.py`

### Model internals

- [Network builders](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/nets.py): `csbdeep/internals/nets.py`
- [Network blocks](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/blocks.py): `csbdeep/internals/blocks.py`
- [Training engine](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/train.py): `csbdeep/internals/train.py`
- [Training losses](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/losses.py): `csbdeep/internals/losses.py`
- [Tiled prediction](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/predict.py): `csbdeep/internals/predict.py`
- [Probabilistic output](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/internals/probability.py): `csbdeep/internals/probability.py`

### Input and interfaces

- [Image file I/O](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/io/__init__.py): `csbdeep/io/__init__.py`
- [Prediction CLI](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/scripts/care_predict.py): `csbdeep/scripts/care_predict.py`
- [Restored images](https://github.com/csbdeep/csbdeep/blob/main/csbdeep/io/__init__.py): `csbdeep/io/__init__.py`

### Other components

- **Microscopy user**

## Connections

- Microscopy user → Raw image pairs: provides images
- Raw image pairs → Patch generation: supplies image pairs
- Patch generation → Image transforms: applies transforms (evidence: `csbdeep/data/generate.py`)
- Patch generation → Image file I/O: saves patches (evidence: `csbdeep/data/generate.py`)
- Microscopy user → Model configuration: configures model
- Model configuration → Model lifecycle: configures lifecycle (evidence: `csbdeep/models/base_model.py`)
- Model lifecycle → Pretrained models: uses pretrained support (evidence: `csbdeep/models/base_model.py`)
- CARE training and prediction → Model lifecycle: extends lifecycle (evidence: `csbdeep/models/care_standard.py`)
- CARE training and prediction → Model configuration: uses configuration (evidence: `csbdeep/models/care_standard.py`)
- CARE training and prediction → Network builders: builds network (evidence: `csbdeep/models/care_standard.py`)
- CARE training and prediction → Training engine: trains with (evidence: `csbdeep/models/care_standard.py`)
- CARE training and prediction → Tiled prediction: predicts with (evidence: `csbdeep/models/care_standard.py`)
- CARE training and prediction → Probabilistic output: returns distributions (evidence: `csbdeep/models/care_standard.py`)
- Network builders → Network blocks: uses blocks (evidence: `csbdeep/internals/nets.py`)
- Training engine → Training losses: uses losses (evidence: `csbdeep/internals/train.py`)
- Isotropic CARE → CARE training and prediction: extends CARE (evidence: `csbdeep/models/care_isotropic.py`)
- Isotropic CARE → Tiled prediction: uses prediction (evidence: `csbdeep/models/care_isotropic.py`)
- Isotropic CARE → Probabilistic output: uses distributions (evidence: `csbdeep/models/care_isotropic.py`)
- Projection CARE → CARE training and prediction: extends CARE (evidence: `csbdeep/models/care_projection.py`)
- Projection CARE → Network builders: uses network builder (evidence: `csbdeep/models/care_projection.py`)
- Projection CARE → Tiled prediction: uses prediction (evidence: `csbdeep/models/care_projection.py`)
- Upsampling CARE → CARE training and prediction: extends CARE (evidence: `csbdeep/models/care_upsampling.py`)
- Prediction CLI → Image file I/O: reads and writes images (evidence: `csbdeep/scripts/care_predict.py`)
- Prediction CLI → Restored images: writes results (evidence: `csbdeep/scripts/care_predict.py`)
- Prediction CLI → CARE training and prediction: invokes model (evidence: `csbdeep/scripts/care_predict.py`)
- Prediction CLI → Data preparation: loads data
- Tiled prediction → Restored images: produces predictions
