Approximating Human Judgment of Generated Image Quality
This addresses the need for efficient and accurate evaluation of generative model outputs, which is crucial for researchers and developers in AI and computer vision, though it is incremental as it builds on existing embedding techniques.
The paper tackles the problem of evaluating image quality from generative models by introducing a method that predicts human judgments of image realism with 66% accuracy, matching human inter-rater agreement, and generalizes across models.
Generative models have made immense progress in recent years, particularly in their ability to generate high quality images. However, that quality has been difficult to evaluate rigorously, with evaluation dominated by heuristic approaches that do not correlate well with human judgment, such as the Inception Score and Fréchet Inception Distance. Real human labels have also been used in evaluation, but are inefficient and expensive to collect for each image. Here, we present a novel method to automatically evaluate images based on their quality as perceived by humans. By not only generating image embeddings from Inception network activations and comparing them to the activations for real images, of which other methods perform a variant, but also regressing the activation statistics to match gold standard human labels, we demonstrate 66% accuracy in predicting human scores of image realism, matching the human inter-rater agreement rate. Our approach also generalizes across generative models, suggesting the potential for capturing a model-agnostic measure of image quality. We open source our dataset of human labels for the advancement of research and techniques in this area.