How cwindolf/dartsort works
Main System Components
The system architecture should be organized into these primary components:
a. Core Processing Pipeline
• Detection Module (src/dartsort/detect)
• Localization Module (src/dartsort/localize)
• Peeling Module (src/dartsort/peel)
• Clustering Module (src/dartsort/cluster)
• Templates Module (src/dartsort/templates)
• Transform Module (src/dartsort/transform)
b. Support Systems
• Utilities (src/dartsort/util)
• Visualization (src/dartsort/vis)
• Configuration (src/dartsort/config.py)
• Pretrained Models (src/dartsort/pretrained)
Data Flow Architecture
Raw Neural Data
• ↓
Detection (spike detection)
• ↓
Localization (spatial positioning)
• ↓
Peeling (signal separation)
• ↓
Clustering (spike sorting)
• ↓
Template Generation
• ↓
Transformed Output
Key Design Elements
Use these visual elements:
• Hexagons for pretrained models
• Diamonds for decision points
• Blue: Core processing modules
• Green: Data input/output
• Yellow: Support utilities
• Purple: Visualization components
Integration Points
Connections with:
• PyTorch integration (localize_torch.py)
• DREDge motion correction
• External data formats
• Visualization pipeline
Technical Details
• GPU acceleration support (indicated by PyTorch usage)
• Parallel processing capabilities (multiprocessing_util.py)
• Model storage (pretrained directory)
• Testing framework integration
Specific Implementation Notes
a. Interface Layer
• CLI interface (cli.py)
• Pipeline scripts
b. Processing Layer
• Core algorithms
• Mathematical operations
c. Infrastructure Layer
• Data utilities
• Configuration management
• Testing framework