How living-technologies/orgasegment works
Main Components
a) Input Layer:
• Image Input Processing (support for JPEG/JPG)
• Configuration Management (from ./conf/)
• Model Loading System
b) Core Processing Components:
• Mask R-CNN Implementation
• Tracking System (based on trackpy)
• Inference Pipeline
• Training Pipeline
c) Interface Layer:
• Streamlit Web Application
• SLURM Batch Processing Interface
Specific Implementation Details
• Configuration Management (OrganoidBasicConfig, app.conf, logging.conf)
• Model Architecture (Mask R-CNN)
• Data Pipeline (train/val/eval splits)
• Tracking System (trackpy integration)
Key Elements to Highlight
a) Two Main Operation Modes:
• Interactive Mode (Streamlit App)
• Batch Processing Mode (SLURM)
b) Data Processing Pipeline:
• Image Input → Preprocessing → Model Inference → Tracking → Results
c) Training Pipeline:
• Dataset → Training → Validation → Evaluation
External Dependencies
• Connections to external systems:
• SLURM Cluster
• GPU Resources
• File System
• Conda/Mamba Environment
Technical Details to Note
Annotations for:
• GPU requirements (>15GB RAM for training)
• Supported platforms (Linux/Windows)
• File format support (JPEG/JPG)
• Python version constraints
Optional Components
Future integration points:
• TensorFlow 2 support (planned)
• Neptune model monitoring (planned)
• Python >3.6 support (planned)