AN AUTOMATED COMPUTER VISION-BASED SYSTEM AND METHOD FOR DETECTION OF TUBERCULOSIS

Abstract

ABSTRACT AN AUTOMATED COMPUTER VISION-BASED SYSTEM AND METHOD FOR DETECTION OF TUBERCULOSIS An automated computer vision-based system and method for detection of tuberculosis includes a network based devices, a server, a network and one or more processors. A data pre-processing unit is configured to process a captured one or more input smear microscopy sputum sample images received from a source machine. A data annotation unit is configured for annotation of received sputum smear microscopy sample images through labelImg 1.4.0 tool. YOLOv7_Lite architecture or model is designed as channel-pruned and scaled-down version that employs SiLU activation, being configured to adjust the depth and to reduce width of the model to enhance detection accuracy of diverse range of features related to bacilli from the received annotated microscopy sputum sample images of tuberculosis and communicating to destination device. The YOLOv7_Lite is computationally lighter deep convolutional model designed or tailored for integration into memory-constrained edge devices and reduction in size is achieved through compound scaling approach.

Specification

Description:Attached Separately , Claims:Attached Separately

Documents