Manish Sharma

Latent representations from a convolutional autoencoder can reveal perceptually meaningful image differences that traditional metrics miss, achieving 98.4% accuracy in distinguishing visually distinct pairs.