# karthikram-p/marine-debris-detection architecture

> Architecture diagram of the GitHub repository karthikram-p/marine-debris-detection, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-08). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/karthikram-p/marine-debris-detection
- Repository: https://github.com/karthikram-p/marine-debris-detection

## Overview

AquaScan is a marine-debris semantic-segmentation project built around Sentinel-2 imagery and UNet++. The principal user workflow is uploading an image in the Streamlit app, preprocessing it, running the model, and downloading a GeoTIFF mask. Separate training and evaluation workflows use MARIDA patches, augmentation, custom loss, metrics, and saved checkpoints. The graph includes documented capabilities, but some wiring—especially app internals and unsampled training/evaluation code—is represented only where the README or source index supports it.

Main entry points: [`semantic_segmentation/unet/train.py`](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/train.py) (Training loop), [`semantic_segmentation/unet/evaluation.py`](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/evaluation.py) (Evaluation runner), [`semantic_segmentation/unet/dataloader.py`](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/dataloader.py) (MARIDA loader), [`semantic_segmentation/unet/app.py`](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/app.py) (Streamlit app)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_inference["Web inference"]
  node_app["Streamlit app<br/>[app.py]"]
  node_modelpp["UNet++ model<br/>[unet_plus_plus.py]"]
  node_mask["Prediction mask"]
end

subgraph group_learning["Training"]
  node_trainer["Training loop<br/>[train.py]"]
  node_loss["Dice loss<br/>[loss_functions.py]"]
  node_checkpoints[("Model checkpoints")]
end

subgraph group_assessment["Evaluation"]
  node_evaluator["Evaluation runner<br/>[evaluation.py]"]
  node_unet["U-Net model<br/>[unet.py]"]
  node_metrics["Segmentation metrics<br/>[metrics.py]"]
  node_predictions["Saved predictions"]
end

subgraph group_domain["Dataset and domain"]
  node_dataset["MARIDA loader<br/>[dataloader.py]"]
  node_patches[("Satellite patches")]
  node_labels["Class mappings<br/>[assets.py]"]
end

node_user(("Image uploader"))

node_user -->|"uploads image"| node_app
node_app -->|"runs inference"| node_modelpp
node_app -->|"writes GeoTIFF"| node_mask
node_mask -->|"downloads mask"| node_user
node_trainer -->|"loads batches"| node_dataset
node_dataset -->|"reads imagery"| node_patches
node_dataset -.->|"maps classes"| node_labels
node_trainer -->|"trains model"| node_modelpp
node_trainer -->|"computes loss"| node_loss
node_trainer -->|"scores model"| node_metrics
node_trainer -->|"saves weights"| node_checkpoints
node_evaluator -->|"loads test set"| node_dataset
node_evaluator -->|"runs evaluation"| node_unet
node_evaluator -->|"calculates scores"| node_metrics
node_evaluator -->|"writes masks"| node_predictions

click node_app "https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/app.py"
click node_modelpp "https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/unet_plus_plus.py"
click node_trainer "https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/train.py"
click node_dataset "https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/dataloader.py"
click node_patches "https://github.com/karthikram-p/marine-debris-detection/tree/main/data/patches"
click node_labels "https://github.com/karthikram-p/marine-debris-detection/blob/main/utils/assets.py"
click node_loss "https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/loss_functions.py"
click node_checkpoints "https://github.com/karthikram-p/marine-debris-detection/tree/main/semantic_segmentation/unet/trained_models"
click node_evaluator "https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/evaluation.py"
click node_unet "https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/unet.py"
click node_metrics "https://github.com/karthikram-p/marine-debris-detection/blob/main/utils/metrics.py"
click node_predictions "https://github.com/karthikram-p/marine-debris-detection/tree/main/data/predicted_unet"

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_app,node_modelpp,node_mask toneBlue
class node_trainer,node_loss,node_checkpoints toneAmber
class node_evaluator,node_unet,node_metrics,node_predictions toneMint
class node_dataset,node_patches,node_labels toneRose
class node_user toneIndigo
```

## Components

### Web inference

- [Streamlit app](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/app.py): `semantic_segmentation/unet/app.py`
- [UNet++ model](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/unet_plus_plus.py): `semantic_segmentation/unet/unet_plus_plus.py`
- **Prediction mask**

### Training

- [Training loop](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/train.py): `semantic_segmentation/unet/train.py`
- [Dice loss](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/loss_functions.py): `semantic_segmentation/unet/loss_functions.py`
- [Model checkpoints](https://github.com/karthikram-p/marine-debris-detection/tree/main/semantic_segmentation/unet/trained_models): `semantic_segmentation/unet/trained_models`

### Evaluation

- [Evaluation runner](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/evaluation.py): `semantic_segmentation/unet/evaluation.py`
- [U-Net model](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/unet.py): `semantic_segmentation/unet/unet.py`
- [Segmentation metrics](https://github.com/karthikram-p/marine-debris-detection/blob/main/utils/metrics.py): `utils/metrics.py`
- [Saved predictions](https://github.com/karthikram-p/marine-debris-detection/tree/main/data/predicted_unet): `data/predicted_unet`

### Dataset and domain

- [MARIDA loader](https://github.com/karthikram-p/marine-debris-detection/blob/main/semantic_segmentation/unet/dataloader.py): `semantic_segmentation/unet/dataloader.py`
- [Satellite patches](https://github.com/karthikram-p/marine-debris-detection/tree/main/data/patches): `data/patches`
- [Class mappings](https://github.com/karthikram-p/marine-debris-detection/blob/main/utils/assets.py): `utils/assets.py`

### Other components

- **Image uploader**

## Connections

- Image uploader → Streamlit app: uploads image (evidence: `README.md`)
- Streamlit app → UNet++ model: runs inference (evidence: `semantic_segmentation/unet/app.py`)
- Streamlit app → Prediction mask: writes GeoTIFF (evidence: `semantic_segmentation/unet/app.py`)
- Prediction mask → Image uploader: downloads mask
- Training loop → MARIDA loader: loads batches (evidence: `semantic_segmentation/unet/train.py`)
- MARIDA loader → Satellite patches: reads imagery (evidence: `semantic_segmentation/unet/dataloader.py`)
- MARIDA loader → Class mappings: maps classes
- Training loop → UNet++ model: trains model (evidence: `semantic_segmentation/unet/train.py`)
- Training loop → Dice loss: computes loss (evidence: `semantic_segmentation/unet/train.py`)
- Training loop → Segmentation metrics: scores model (evidence: `semantic_segmentation/unet/train.py`)
- Training loop → Model checkpoints: saves weights (evidence: `semantic_segmentation/unet/train.py`)
- Evaluation runner → MARIDA loader: loads test set (evidence: `semantic_segmentation/unet/evaluation.py`)
- Evaluation runner → U-Net model: runs evaluation (evidence: `semantic_segmentation/unet/evaluation.py`)
- Evaluation runner → Segmentation metrics: calculates scores (evidence: `semantic_segmentation/unet/evaluation.py`)
- Evaluation runner → Saved predictions: writes masks (evidence: `semantic_segmentation/unet/evaluation.py`)
