The paper introduces GANomaly, a novel semi-supervised anomaly detection model that leverages conditional generative adversarial networks (GANs) to learn the normal data distribution and identify anomalies. The model uses an encoder-decoder-encoder architecture to map input images to a lower-dimensional vector, which is then used to reconstruct the image. An additional encoder maps the generated image to its latent representation, aiding in learning the data distribution. During training, the model minimizes the distance between the input and reconstructed images, and during inference, a larger distance metric indicates an anomaly. The approach is evaluated on various benchmark datasets, demonstrating superior performance compared to state-of-the-art methods, both statistically and computationally. The key contributions include a novel adversarial autoencoder architecture, efficient training and inference, and reproducibility through publicly available code.The paper introduces GANomaly, a novel semi-supervised anomaly detection model that leverages conditional generative adversarial networks (GANs) to learn the normal data distribution and identify anomalies. The model uses an encoder-decoder-encoder architecture to map input images to a lower-dimensional vector, which is then used to reconstruct the image. An additional encoder maps the generated image to its latent representation, aiding in learning the data distribution. During training, the model minimizes the distance between the input and reconstructed images, and during inference, a larger distance metric indicates an anomaly. The approach is evaluated on various benchmark datasets, demonstrating superior performance compared to state-of-the-art methods, both statistically and computationally. The key contributions include a novel adversarial autoencoder architecture, efficient training and inference, and reproducibility through publicly available code.