RDT-FSDet: Few-shot Object Detection for Rapid Antigen Test

Abstract

Abstract Objective: Manual verification of rapid diagnostic test (RDT) results is a time-consuming task; therefore, it is essential to introduce an object detection model into RDT result recognition to reduce the time involved. To address these problems, a detector that can rapidly adapt to different RDT results in various regions is important. Methods: We employed the few-shot object detection strategy and trained the Faster R-CNN detector with the mainland data set as the base class, followed by fine-tuning with the few-shot approach on the Macau RDT result data set. Moreover, we introduced 2 novel data augmentation methods, namely the “light simulation mask method” and “synthetic positive samples” for an unbalanced data set, to increase the sample size and balance the data set of the RDT detection task. Result: Compared with LightR-YOLOv5, RDT-few-shot detector (FSDet) achieved an mean average precision of 91.18 and a recall of 93.59 on the Macau RDT data set, demonstrating that this model can rapidly adapt to RDT results in different regions. The inference time of RDT-FSDet for each RDT result was 0.14 seconds, which can save ~90% of the detection time compared with manual screening. Conclusion: In addition to its application in the context of the coronavirus disease 2019 pandemic, this model can also be used as a general small-sample detection model. RDT-FSDet can be applied to the detection tasks of other small data sets, such as managing and analyzing detection results in other or future epidemics.