# nvlabs/spade architecture

> Architecture diagram of the GitHub repository nvlabs/spade, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-01). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/nvlabs/spade
- Repository: https://github.com/nvlabs/spade

## Overview

SPADE is a research implementation for semantic image synthesis: it turns paired semantic labels (and optional instance maps) into generated images, supporting dataset preparation, model training, and checkpoint-based inference. Training is initiated by the training entry point, which loads data and delegates optimization to the trainer; testing uses the model to generate result images. Dataset variants, SPADE network components, and optional logging/output capabilities are represented. Edges follow shown source references or README statements; unsampled entry-point details are limited to documented relationships.

Main entry points: [`train.py`](https://github.com/nvlabs/spade/blob/master/train.py) (Training entry), [`models/pix2pix_model.py`](https://github.com/nvlabs/spade/blob/master/models/pix2pix_model.py) (Pix2Pix model), [`test.py`](https://github.com/nvlabs/spade/blob/master/test.py) (Testing entry), [`util/visualizer.py`](https://github.com/nvlabs/spade/blob/master/util/visualizer.py) (Progress output)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_workflow["Run workflow"]
  node_train_entry["Training entry<br/>[train.py]"]
  node_test_entry["Testing entry<br/>[test.py]"]
  node_options["Run options<br/>[base_options.py]"]
  node_trainer["Training harness<br/>[pix2pix_trainer.py]"]
end

subgraph group_data["Semantic data"]
  node_dataset_registry["Dataset selection<br/>[__init__.py]"]
  node_dataset_adapters["Dataset adapters<br/>[coco_dataset.py]"]
  node_paired_dataset["Paired samples<br/>[pix2pix_dataset.py]"]
  node_image_folder["Image folder<br/>[image_folder.py]"]
  node_preprocessing["Image transforms<br/>[base_dataset.py]"]
end

subgraph group_model["Synthesis model"]
  node_pix2pix_model["Pix2Pix model<br/>[pix2pix_model.py]"]
  node_generator["SPADE generator<br/>[generator.py]"]
  node_discriminator["Discriminator<br/>[discriminator.py]"]
  node_encoder["VAE encoder<br/>[encoder.py]"]
  node_architecture["Network blocks<br/>[architecture.py]"]
  node_normalization["SPADE normalization<br/>[normalization.py]"]
  node_losses["GAN losses<br/>[loss.py]"]
end

subgraph group_support["Training support"]
  node_instance_maps["Instance maps"]
  node_iteration["Iteration tracking<br/>[iter_counter.py]"]
  node_visualizer["Progress output<br/>[visualizer.py]"]
  node_image_utils["Image utilities<br/>[util.py]"]
  node_html["HTML results<br/>[html.py]"]
end

node_researcher(("Researcher"))
node_tensorboard["TensorBoard"]
node_image_results["Generated images"]

node_researcher -->|"starts training"| node_train_entry
node_researcher -->|"starts inference"| node_test_entry
node_train_entry -->|"reads options"| node_options
node_train_entry -->|"loads dataset"| node_dataset_registry
node_train_entry -->|"runs training"| node_trainer
node_train_entry -->|"tracks epochs"| node_iteration
node_train_entry -->|"reports progress"| node_visualizer
node_test_entry -->|"loads dataset"| node_dataset_registry
node_test_entry -->|"generates images"| node_pix2pix_model
node_test_entry -->|"writes results"| node_image_results
node_options -->|"configures dataset"| node_dataset_registry
node_dataset_registry -->|"selects adapter"| node_dataset_adapters
node_dataset_adapters -->|"extends paired loader"| node_paired_dataset
node_paired_dataset -->|"transforms samples"| node_preprocessing
node_paired_dataset -->|"discovers images"| node_image_folder
node_instance_maps -->|"prepares COCO maps"| node_dataset_adapters
node_trainer -->|"optimizes model"| node_pix2pix_model
node_pix2pix_model -->|"runs generator"| node_generator
node_pix2pix_model -->|"runs discriminator"| node_discriminator
node_pix2pix_model -->|"encodes VAE input"| node_encoder
node_pix2pix_model -->|"computes GAN loss"| node_losses
node_generator -->|"uses blocks"| node_architecture
node_architecture -->|"applies SPADE"| node_normalization
node_visualizer -->|"converts visuals"| node_image_utils
node_visualizer -->|"renders galleries"| node_html
node_visualizer -.->|"logs summaries"| node_tensorboard

click node_train_entry "https://github.com/nvlabs/spade/blob/master/train.py"
click node_test_entry "https://github.com/nvlabs/spade/blob/master/test.py"
click node_options "https://github.com/nvlabs/spade/blob/master/options/base_options.py"
click node_trainer "https://github.com/nvlabs/spade/blob/master/trainers/pix2pix_trainer.py"
click node_dataset_registry "https://github.com/nvlabs/spade/blob/master/data/__init__.py"
click node_dataset_adapters "https://github.com/nvlabs/spade/blob/master/data/coco_dataset.py"
click node_paired_dataset "https://github.com/nvlabs/spade/blob/master/data/pix2pix_dataset.py"
click node_image_folder "https://github.com/nvlabs/spade/blob/master/data/image_folder.py"
click node_preprocessing "https://github.com/nvlabs/spade/blob/master/data/base_dataset.py"
click node_instance_maps "https://github.com/nvlabs/spade/blob/master/datasets/coco_generate_instance_map.py"
click node_pix2pix_model "https://github.com/nvlabs/spade/blob/master/models/pix2pix_model.py"
click node_generator "https://github.com/nvlabs/spade/blob/master/models/networks/generator.py"
click node_discriminator "https://github.com/nvlabs/spade/blob/master/models/networks/discriminator.py"
click node_encoder "https://github.com/nvlabs/spade/blob/master/models/networks/encoder.py"
click node_architecture "https://github.com/nvlabs/spade/blob/master/models/networks/architecture.py"
click node_normalization "https://github.com/nvlabs/spade/blob/master/models/networks/normalization.py"
click node_losses "https://github.com/nvlabs/spade/blob/master/models/networks/loss.py"
click node_iteration "https://github.com/nvlabs/spade/blob/master/util/iter_counter.py"
click node_visualizer "https://github.com/nvlabs/spade/blob/master/util/visualizer.py"
click node_image_utils "https://github.com/nvlabs/spade/blob/master/util/util.py"
click node_html "https://github.com/nvlabs/spade/blob/master/util/html.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_train_entry,node_test_entry,node_options,node_trainer toneBlue
class node_dataset_registry,node_dataset_adapters,node_paired_dataset,node_image_folder,node_preprocessing toneAmber
class node_pix2pix_model,node_generator,node_discriminator,node_encoder,node_architecture,node_normalization,node_losses toneMint
class node_instance_maps,node_iteration,node_visualizer,node_image_utils,node_html toneRose
class node_researcher,node_tensorboard,node_image_results toneIndigo
```

## Components

### Run workflow

- [Training entry](https://github.com/nvlabs/spade/blob/master/train.py): `train.py`
- [Testing entry](https://github.com/nvlabs/spade/blob/master/test.py): `test.py`
- [Run options](https://github.com/nvlabs/spade/blob/master/options/base_options.py): `options/base_options.py`
- [Training harness](https://github.com/nvlabs/spade/blob/master/trainers/pix2pix_trainer.py): `trainers/pix2pix_trainer.py`

### Semantic data

- [Dataset selection](https://github.com/nvlabs/spade/blob/master/data/__init__.py): `data/__init__.py`
- [Dataset adapters](https://github.com/nvlabs/spade/blob/master/data/coco_dataset.py): `data/coco_dataset.py`
- [Paired samples](https://github.com/nvlabs/spade/blob/master/data/pix2pix_dataset.py): `data/pix2pix_dataset.py`
- [Image folder](https://github.com/nvlabs/spade/blob/master/data/image_folder.py): `data/image_folder.py`
- [Image transforms](https://github.com/nvlabs/spade/blob/master/data/base_dataset.py): `data/base_dataset.py`

### Synthesis model

- [Pix2Pix model](https://github.com/nvlabs/spade/blob/master/models/pix2pix_model.py): `models/pix2pix_model.py`
- [SPADE generator](https://github.com/nvlabs/spade/blob/master/models/networks/generator.py): `models/networks/generator.py`
- [Discriminator](https://github.com/nvlabs/spade/blob/master/models/networks/discriminator.py): `models/networks/discriminator.py`
- [VAE encoder](https://github.com/nvlabs/spade/blob/master/models/networks/encoder.py): `models/networks/encoder.py`
- [Network blocks](https://github.com/nvlabs/spade/blob/master/models/networks/architecture.py): `models/networks/architecture.py`
- [SPADE normalization](https://github.com/nvlabs/spade/blob/master/models/networks/normalization.py): `models/networks/normalization.py`
- [GAN losses](https://github.com/nvlabs/spade/blob/master/models/networks/loss.py): `models/networks/loss.py`

### Training support

- [Instance maps](https://github.com/nvlabs/spade/blob/master/datasets/coco_generate_instance_map.py): `datasets/coco_generate_instance_map.py`
- [Iteration tracking](https://github.com/nvlabs/spade/blob/master/util/iter_counter.py): `util/iter_counter.py`
- [Progress output](https://github.com/nvlabs/spade/blob/master/util/visualizer.py): `util/visualizer.py`
- [Image utilities](https://github.com/nvlabs/spade/blob/master/util/util.py): `util/util.py`
- [HTML results](https://github.com/nvlabs/spade/blob/master/util/html.py): `util/html.py`

### Other components

- **Researcher**
- **TensorBoard**
- **Generated images**

## Connections

- Researcher → Training entry: starts training (evidence: `README.md`)
- Researcher → Testing entry: starts inference (evidence: `README.md`)
- Training entry → Run options: reads options (evidence: `README.md`)
- Training entry → Dataset selection: loads dataset (evidence: `README.md`)
- Training entry → Training harness: runs training (evidence: `README.md`)
- Training entry → Iteration tracking: tracks epochs (evidence: `README.md`)
- Training entry → Progress output: reports progress (evidence: `README.md`)
- Testing entry → Dataset selection: loads dataset (evidence: `README.md`)
- Testing entry → Pix2Pix model: generates images (evidence: `README.md`)
- Testing entry → Generated images: writes results (evidence: `README.md`)
- Run options → Dataset selection: configures dataset (evidence: `options/base_options.py`)
- Dataset selection → Dataset adapters: selects adapter (evidence: `data/__init__.py`)
- Dataset adapters → Paired samples: extends paired loader (evidence: `data/coco_dataset.py`)
- Paired samples → Image transforms: transforms samples (evidence: `data/pix2pix_dataset.py`)
- Paired samples → Image folder: discovers images (evidence: `data/coco_dataset.py`)
- Instance maps → Dataset adapters: prepares COCO maps (evidence: `README.md`)
- Training harness → Pix2Pix model: optimizes model (evidence: `trainers/pix2pix_trainer.py`)
- Pix2Pix model → SPADE generator: runs generator (evidence: `models/pix2pix_model.py`)
- Pix2Pix model → Discriminator: runs discriminator (evidence: `models/pix2pix_model.py`)
- Pix2Pix model → VAE encoder: encodes VAE input (evidence: `models/pix2pix_model.py`)
- Pix2Pix model → GAN losses: computes GAN loss (evidence: `models/pix2pix_model.py`)
- SPADE generator → Network blocks: uses blocks (evidence: `models/networks/generator.py`)
- Network blocks → SPADE normalization: applies SPADE (evidence: `models/networks/architecture.py`)
- Progress output → Image utilities: converts visuals (evidence: `util/visualizer.py`)
- Progress output → HTML results: renders galleries (evidence: `util/visualizer.py`)
- Progress output → TensorBoard: logs summaries (evidence: `util/visualizer.py`)
