Variational Autoencoder for Anomaly Detection: A Comparative Study
This is an incremental study that addresses the problem of evaluating anomaly detection methods for researchers by providing a comparative analysis with new benchmarking.
This paper compared Variational Autoencoder architectures for anomaly detection, finding that ViT-VAE performed best across scenarios while VAE-GRF required more tuning, and used the MiAD dataset to reduce over-reliance on MVTec for more robust evaluation.
This paper aims to conduct a comparative analysis of contemporary Variational Autoencoder (VAE) architectures employed in anomaly detection, elucidating their performance and behavioral characteristics within this specific task. The architectural configurations under consideration encompass the original VAE baseline, the VAE with a Gaussian Random Field prior (VAE-GRF), and the VAE incorporating a vision transformer (ViT-VAE). The findings reveal that ViT-VAE exhibits exemplary performance across various scenarios, whereas VAE-GRF may necessitate more intricate hyperparameter tuning to attain its optimal performance state. Additionally, to mitigate the propensity for over-reliance on results derived from the widely used MVTec dataset, this paper leverages the recently-public MiAD dataset for benchmarking. This deliberate inclusion seeks to enhance result competitiveness by alleviating the impact of domain-specific models tailored exclusively for MVTec, thereby contributing to a more robust evaluation framework. Codes is available at https://github.com/endtheme123/VAE-compare.git.