Early Grayscale Training Enhances Vision in AI/Humans
Early Grayscale Training Enhances Vision in AI/Humans
Impact of Early Visual Experience On Later Usage of Color Cues
Tags: IIT Delhi, India, Computing Technology, Healthcare & Lifesciences
Research shows that normally sighted children can recognize objects equally well in color and grayscale images. However, newly sighted children with congenital cataract, known as Prakash children, perform worse with grayscale images, likely due to an overreliance on color information developed after surgery. Computational models support this hypothesis, demonstrating that networks trained first with grayscale images perform better with color variations than those trained first with color images. Applications of this technology include improving AI vision systems by mimicking human visual development patterns. The findings also provide insights into visual deficits and normal visual development.
IP Type or Form Factor: Discovery & Research; Software & Algorithm
TRL: 3 - proof of concept with needs validated
Industry or Tech Area: Big Data Analytics & Simulations; Healthcare Provider
