How walkerad03/sight-seer works
Main Components
a) Data Pipeline Layer:
• "preprocessing" module as the entry point
• Two sub-components:
• Image Downloader (download_images.py)
• Dataset Builder (build_dataset.py)
b) ML Model Kit Layer:
• Core ML functionality in "ml_model_kit"
• Key components:
• Data Setup (data_setup.py)
• Model Builder (model_builder.py)
• Training Engine (engine.py)
• Utilities (utils.py)
c) Inference Layer:
• Main.py as the orchestrator
• Prediction flow
Data Flow (top to bottom)
• Raw Images → Preprocessing → Structured Dataset → ML Pipeline → Trained Model → Inference
Component Interaction Flow
• main.py ↔ ml_model_kit components ↔ preprocessing components
Detailed Component Breakdown
Preprocessing Module
• Image Downloader
• Dataset Builder
ML Model Kit
• Data Setup
• Model Builder
• Training Engine
• Utilities
Additional Elements
Technical details:
• Input: Single photograph
• Output: Geographical coordinates
• Model type: ResNet
• Data flow direction
• Training/inference paths
Remember to:
• Keep the ResNet architecture as a central focus
• Clear separation between preprocessing and model operations
• Data transformation points
• Error handling flows
• Validation steps