Visual defect detection is critical to ensure the quality of most products. However, majority of small medium manufactures still rely on tedious and error-prune human manual inspection. The main reasons include: 1) the existing automated visual defect detection systems require altering production assembly lines, which is time consuming and expensive 2) the existing systems require manually collecting defective samples and labeling them for a comparison-based algorithm or training a machine learning model. This introduces heavy burden for Small and Medium-sized Enterprise (SME) manufactures as defects do not happen often and are difficult and time-consuming to collect. Furthermore, we cannot exhaustively collect or define all defect types as any new deviation from acceptable products are defects. In this paper, we overcome these challenges and design a three-stage plug-and-play fully automated unsupervised 360-degree defect detection system. In our system, products are freely placed on an unaltered assembly line and receive 360 degree visual inspection with multiple cameras from different angles. As such, the images collected from real-world product assembly lines contain lots of background noise. The products face different angles. The product sizes vary due to the distance to cameras. All these make defect detection much more difficult. Our system use object detection, background subtraction and unsupervised normalizing flow-based defect detection techniques to tackle these difficulty. Experiments show our system can achieve 0.90 AUROC in a real-world non-altered drink ware production assembly line.