COVID-19 detection based on Computer Vision and Big Data

Abstract

For detecting COVID-19 and checking the severity of the patient's condition, CT examination of the lungs is significant. However, the current manual viewing of CT images requires professionalism. In order to improve the inspection efficiency of the huge number of CT images, it is necessary to develop an intelligent detection algorithm to perform CT inspections. This paper proposes a COVID-19 detection algorithm based on EfficientDet. EfficientDet leverages a faster and easier multi-scale fusion approach, which is more suitable for COVID-19 detection tasks with finer feature granularity. In addition, data augmentation is also significant in COVID-19 detection tasks. This paper verifies the effectiveness of EfficientDet on the SIIMFISABIO-RSNA COVID-19 Detection dataset provided by Kaggle platform. Experimental results show that EfficientDet has achieved better performance than other detection algorithms. Taking MAP@0.5 as an indicator, EfficientDet reaches 0.545, which is 7.9% and 3.3% higher than the Faster RCNN algorithm and YOLO-V5.

Ähnliche Arbeiten