RESUMEN
In numerous applications, abnormal samples are hard to collect, limiting the use of well-established supervised learning methods. GAN-based models which trained in an unsupervised and single feature set manner have been proposed by simultaneously considering the reconstruction error and the latent space deviation between normal samples and abnormal samples. However, the ability to capture the input distribution of each feature set is limited. Hence, we propose an unsupervised and multi-feature model, Wave-GANomaly, trained only on normal samples to learn the distribution of these normal samples. The model predicts whether a given sample is normal or not by its deviation from the distribution of normal samples. Wave-GANomaly fuses and selects from the wave-based features extracted by the WaveBlock module and the convolution-based features. The WaveBlock has proven to efficiently improve the performance on image classification, object detection, and segmentation tasks. As a result, Wave-GANomaly achieves the best average area under the curve (AUC) on the Canadian Institute for Advanced Research (CIFAR)-10 dataset (94.3%) and on the Modified National Institute of Standards and Technology (MNIST) dataset (91.0%) when compared to existing state-of-the-art anomaly detectors such as GANomaly, Skip-GANomaly, and the skip-attention generative adversarial network (SAGAN). We further verify our method by the self-curated real-world dataset, the result show that our method is better than GANomaly which only use single feature set for training the model.
RESUMEN
In electronics manufacturing, surface defect detection is very important for product quality control, and defective products can cause severe customer complaints. At the same time, in the manufacturing process, the cycle time of each product is usually very short. Furthermore, high-resolution input images from high-resolution industrial cameras are necessary to meet the requirements for high quality control standards. Hence, how to design an accurate object detector with real-time inference speed that can accept high-resolution input is an important task. In this work, an accurate YOLO-style object detector was designed, ATT-YOLO, which uses only one self-attention module, many-scale feature extraction and integration in the backbone and feature pyramid, and an improved auto-anchor design to address this problem. There are few datasets for surface detection in electronics manufacturing. Hence, we curated a dataset consisting of 14,478 laptop surface defects, on which ATT-YOLO achieved 92.8% mAP0.5 for the binary-class object detection task. We also further verified our design on the COCO benchmark dataset. Considering both computation costs and the performance of object detectors, ATT-YOLO outperforms several state-of-the-art and lightweight object detectors on the COCO dataset. It achieves a 44.9% mAP score and 21.8 GFLOPs, which is better than the compared models including YOLOv8-small (44.9%, 28.6G), YOLOv7-tiny-SiLU (38.7%, 13.8G), YOLOv6-small (43.1%, 44.2G), pp-YOLOE-small (42.7%, 17.4G), YOLOX-small (39.6%, 26.8G), and YOLOv5-small (36.7%, 17.2G). We hope that this work can serve as a useful reference for the utilization of attention-based networks in real-world situations.