taeba/TERRA_AI

🌍 TERRA-EYE: Off-Road Terrain Segmentation

This Space hosts the TERRA-EYE semantic segmentation demo using a DeepLabV3+ (ResNet50) model trained on the Duality AI challenge dataset.

How it works

1. 1.Inference Backend: A FastAPI service handles the PyTorch model inference on CPU.
2. 2.Frontend UI: A Streamlit dashboard allows for image uploads and visualization.

Features

• —Side-by-side comparison.
• —Interactive alpha-blending slider.
• —Class distribution analytics.
• —High-contrast, color-blind friendly palette.

Technical Details

• —Architecture: DeepLabV3+
• —Backbone: ResNet50
• —Environment: CPU-only Production (FastAPI + Streamlit)