Vision Transformer in Industrial Visual Inspection

Artificial intelligence as an approach to visual inspection in industrial applications has been considered for decades. Recent successes, driven by advances deep learning, present a potential paradigm shift and have the facilitate automated inspection, even under complex environmental conditions. Thereby, convolutional neural networks (CNN) de facto standard deep-learning-based computer vision (CV) last 10 years. Recently, attention-based transformer architectures emerged surpassed performance of CNNs on benchmark datasets, regarding regular CV tasks, such image classification, object detection, or segmentation. Nevertheless, despite their outstanding results, application transformers real world is sparse. We suspect that this likely due assumption they require enormous amounts data be effective. In study, we evaluate assumption. For this, perform systematic comparison seven widely-used state-of-the-art CNN based trained three different use cases domain damage assessment railway freight car maintenance. show models achieve at least equivalent with sparse available, significantly surpass them increasingly tasks.