Visual Language Modeling on CNN Image Representations
This addresses the issue of CNN insensitivity to image naturalness for tasks like image generation and unnatural region detection, though it is incremental as it builds on existing CNN and language model techniques.
The paper tackled the problem of measuring image naturalness by proposing a method that uses a variant of Recurrent Neural Network Language Models on pre-trained CNN representations, resulting in more understandable image generation and state-of-the-art eye fixation prediction on two datasets.
Measuring the naturalness of images is important to generate realistic images or to detect unnatural regions in images. Additionally, a method to measure naturalness can be complementary to Convolutional Neural Network (CNN) based features, which are known to be insensitive to the naturalness of images. However, most probabilistic image models have insufficient capability of modeling the complex and abstract naturalness that we feel because they are built directly on raw image pixels. In this work, we assume that naturalness can be measured by the predictability on high-level features during eye movement. Based on this assumption, we propose a novel method to evaluate the naturalness by building a variant of Recurrent Neural Network Language Models on pre-trained CNN representations. Our method is applied to two tasks, demonstrating that 1) using our method as a regularizer enables us to generate more understandable images from image features than existing approaches, and 2) unnaturalness maps produced by our method achieve state-of-the-art eye fixation prediction performance on two well-studied datasets.