AI Learns from Simulated Images, Automates Urban Landscaping
AI Learns from Simulated Images, Automates Urban Landscaping
Images of Simulated Cities Help Artificial Intelligence to Understand Real Streetscapes
Tags: Osaka University, Japan, Construction & Smart Cities, Computing Technology
Researchers have developed a method to train AI for urban landscaping using computer-simulated images, eliminating the need for manual segmentation. Realistic 3D city models generate accurate ground truth data, and a generative adversarial network produces photorealistic street-view images. This innovative approach greatly reduces dataset preparation time and cost. Trained models on simulated data performed comparably to those trained on real data, demonstrating the method's effectiveness. This breakthrough could revolutionize deep learning-assisted urban planning by providing abundant training data with minimal effort.
IP Type or Form Factor: Process & Method; Software & Algorithm
TRL: Not specified
Industry or Tech Area: Smart Cities & Urban Planning; Big Data Analytics & Simulations
