Synthetic-Neuroscore: Using A Neuro-AI Interface for Evaluating Generative Adversarial Networks

TL;DR

This paper introduces Neuroscore, a new evaluation metric for GANs that uses brain signals to better reflect human perception, requiring fewer samples and providing more accurate rankings of image quality.

Contribution

The paper proposes Neuroscore, a brain signal-based evaluation metric for GANs, and a CNN-based neuro-AI interface to predict it directly from generated images, enhancing evaluation accuracy.

Findings

Neuroscore aligns more closely with human judgment than existing metrics.

The neuro-AI interface predicts Neuroscore effectively without neural responses.

Including neural responses during training improves prediction accuracy.

Abstract

Generative adversarial networks (GANs) are increasingly attracting attention in the computer vision, natural language processing, speech synthesis and similar domains. Arguably the most striking results have been in the area of image synthesis. However, evaluating the performance of GANs is still an open and challenging problem. Existing evaluation metrics primarily measure the dissimilarity between real and generated images using automated statistical methods. They often require large sample sizes for evaluation and do not directly reflect human perception of image quality. In this work, we describe an evaluation metric we call Neuroscore, for evaluating the performance of GANs, that more directly reflects psychoperceptual image quality through the utilization of brain signals. Our results show that Neuroscore has superior performance to the current evaluation metrics in that: (1) It is more consistent with human judgment; (2) The evaluation process needs much smaller numbers of samples; and (3) It is able to rank the quality of images on a per GAN basis. A convolutional neural network (CNN) based neuro-AI interface is proposed to predict Neuroscore from GAN-generated images directly without the need for neural responses. Importantly, we show that including neural responses during the training phase of the network can significantly improve the prediction capability of the proposed model. Materials related to this work are provided at https://github.com/villawang/Neuro-AI-Interface.