Flood Area Segmentation System

My Role

• 01Trained the flood segmentation models and prepared the inference workflow.
• 02Built the backend pipeline for image upload, mask generation, and comparison output.
• 03Deployed the inference system on Microsoft Azure for online access.

Features

• 01Automatic flood segmentation from uploaded images to generate clear flood masks.
• 02Two models in one system: U-Net and U-Net++ for side-by-side comparison.
• 03Analysis statistics: flood area, flood pixel count, total pixels, and model difference summary.
• 04Compare mode showing original image, model masks, and disagreement map.
• 05Model agreement score to measure prediction consistency.
• 06Python backend inference pipeline deployed to Microsoft Azure.

Impact

• Enables rapid identification of flood-affected areas from imagery.
• Provides quantitative estimates that support reporting and monitoring.
• Facilitates segmentation model evaluation through direct model comparison.

Stack