TID2013

New autoencoder method quantifies image differences using latent space

Researchers have developed a new method for quantifying image differences by utilizing latent representations learned by deep neural networks. This approach uses a convolutional autoencoder to compute cosine similarity …