Low-Quality Image Detection by Hierarchical VAE
This addresses the need for automated filtering of low-quality images in applications like employee rosters or training datasets, though it is incremental as it builds on existing VAE techniques.
The study tackled the problem of unsupervised detection of low-quality images by proposing a method that uses hierarchical variational autoencoders to identify degraded images and provide visual clues, achieving superior performance over other unsupervised out-of-distribution detection methods.
To make an employee roster, photo album, or training dataset of generative models, one needs to collect high-quality images while dismissing low-quality ones. This study addresses a new task of unsupervised detection of low-quality images. We propose a method that not only detects low-quality images with various types of degradation but also provides visual clues of them based on an observation that partial reconstruction by hierarchical variational autoencoders fails for low-quality images. The experiments show that our method outperforms several unsupervised out-of-distribution detection methods and also gives visual clues for low-quality images that help humans recognize them even in thumbnail view.