Abstract:
Thorax illness is a serious medical problem. Approximately 3 million individuals die each
year from thoracic disorders, according to a recent report [10]. Inspection of medical
images demands strong degree of professional experience and focus. Furthermore, it also
takes a lot of time and costly. As a result, it is critical to automate the diagnosis of
thorax disorders using chest radiography. In this study, we used an ensemble strategy
to detect anomalies in chest radiographs having limited image size of 512x512, we also
utilized two class filters over the predictions of our two models i.e. Yolov5 and faster
R-CNN, and applied augmentations as well. The implemented methodology was tested
using the VinBigData Chest X-ray Abnormalities Detection Competition [11] evaluation
method, and it received one of the top score in the challenge.
