https://github.com/awrsha/coral-reef-research

An integrated suite for coral reef research combining Meta AI's SAM2 and automated annotation tools.
https://github.com/awrsha/coral-reef-research

annotation coral-reefs sam2 segment-anything-model segmentation

Last synced: 3 months ago
JSON representation

An integrated suite for coral reef research combining Meta AI's SAM2 and automated annotation tools.

Host: GitHub
URL: https://github.com/awrsha/coral-reef-research
Owner: Awrsha
License: apache-2.0
Created: 2024-11-20T16:04:28.000Z (12 months ago)
Default Branch: master
Last Pushed: 2025-03-20T13:52:50.000Z (8 months ago)
Last Synced: 2025-04-10T05:41:03.657Z (7 months ago)
Topics: annotation, coral-reefs, sam2, segment-anything-model, segmentation
Language: Jupyter Notebook
Homepage:
Size: 83.9 MB
Stars: 1
Watchers: 2
Forks: 0
Open Issues: 0
Metadata Files:
- Readme: README.md
- License: LICENSE

Awesome Lists containing this project

README

# 🌊 Advanced Coral Reef Analysis with SAM2 & COCO Tools
> An integrated suite for coral reef research combining Meta AI's SAM2 and automated annotation tools

.:: Advanced Coral Reef Analysis with SAM2 & COCO Tools ::.

subgraph "Processing Pipeline"
B
C
D
end

subgraph "Data Format Handling"
E
F
end
```

## 🚀 Key Features

```mermaid
mindmap
root((Coral Analysis Suite))
SAM2 Integration
Advanced instance segmentation
Interactive prompting
Automated mask generation
Format Conversion
CSV to COCO converter
Automatic ID generation
Metadata enrichment
Quality Control
Error handling
Validation checks
Progress tracking
Export Capabilities
COCO JSON format
CSV annotations
Visualization tools
```

## ⚙️ Components

1. **SAM2 Segmentation Engine** - Instance segmentation
- Multi-mask output
- GPU optimization
- Interactive prompting

2. **Format Converter**
- CSV to COCO transformation
- Deterministic ID generation
- Comprehensive error handling
- Progress tracking

3. **Visualization Tools**
- Mask visualization
- Bounding box display
- Annotation overlay
- Quality metrics

## 🛠️ Installation

```bash
# Clone repository
git clone https://github.com/Awrsha/Coral-Reef-Research.git
cd Coral-Reef-Research

# Install dependencies
pip install -r requirements.txt

# Download SAM2 model
wget https://dl.fbaipublicfiles.com/segment_anything_2/072824/sam2_hiera_large.pt -P checkpoints/
```

## 📊 Data Flow
```mermaid
sequenceDiagram
participant User
participant SAM2
participant Converter
participant Storage

User->>SAM2: Input Image
SAM2->>SAM2: Generate Masks
SAM2->>Converter: Segmentation Data
Converter->>Converter: Format Processing
Converter->>Storage: COCO JSON
Storage->>User: Final Dataset
```

## 💻 Usage Examples

### SAM2 Segmentation
```python
from sam2_coral_segmentation import setup_sam2, process_image

# Initialize model
model, device = setup_sam2()

# Process image
masks, scores = process_image("coral_image.jpg", model)
```

### Format Conversion
```python
from coco_converter import COCOConverter

converter = COCOConverter(
input_path="annotations.csv",
output_path="coco_dataset.json",
img_width=2704,
img_height=1524
)

coco_data = converter.convert()
converter.save(coco_data)
```

## 📊 Output Formats

### COCO JSON Structure

```json
{
"info": {"description": "Coral Reef Dataset"},
"images": [
{
"id": 1,
"width": 2704,
"height": 1524,
"file_name": "coral_001.jpg"
}
],
"annotations": [...],
"categories": [...]
}
```

## Developers 👨🏻‍💻