### Install Dependencies with uv

Source: https://github.com/qbammey/synthbuster/blob/main/readme.md

Installs project dependencies using Astral uv. Ensure the repository is cloned before running.

```bash
uv sync
```

--------------------------------

### Example Configuration for Training

Source: https://github.com/qbammey/synthbuster/blob/main/readme.md

A JSON configuration file for training Synthbuster models. Specifies dataset paths, model parameters, and training hyperparameters.

```json
{
  "dataset_root": "/path/to/dataset",
  "file_extension": ".png",
  "method": "rank",
  "rank_sz": 4,
  "max_period": 16,
  "jpeg_mode": "none",
  "learning_rate": 0.05,
  "max_iter": 200,
  "max_depth": 3,
  "max_leaf_nodes": 31,
  "min_samples_leaf": 50,
  "l2_regularization": 1e-6,
  "random_state": 753
}
```

--------------------------------

### Launch Streamlit Demo

Source: https://github.com/qbammey/synthbuster/blob/main/readme.md

Starts the interactive Streamlit demo for Synthbuster. Allows users to choose models, upload images, and see live predictions.

```bash
uv run streamlit run streamlit_app.py
```

--------------------------------

### Train a Model with Fixed Configuration

Source: https://github.com/qbammey/synthbuster/blob/main/readme.md

Trains a Synthbuster model using a specified configuration file. Saves the trained model and its configuration.

```bash
uv run train_fixed.py \
  --config config.json \
  --save-model models/model.joblib \
  --save-config models/config.json
```

--------------------------------

### Inference on a Single Image

Source: https://github.com/qbammey/synthbuster/blob/main/readme.md

Runs a trained Synthbuster model on a single image to predict its probability of being synthetic and the class decision.

```bash
uv run infer_image.py \
  --model models/model.joblib \
  --config models/config.json \
  --image data/test/real/example.png
```

--------------------------------

### Evaluate Model on a Dataset Split

Source: https://github.com/qbammey/synthbuster/blob/main/readme.md

Evaluates a trained Synthbuster model on a specified dataset split (train, val, test, or all). Results are saved to a JSON file.

```bash
uv run evaluate_dataset.py \
  --model models/model.joblib \
  --config models/config.json \
  --split test \
  --n-jobs -1 \
  --results results_test.json
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.