ANALYSIS OF WEAPON AND FIRE DETECTION ON DIFFERENT DATASETS IN ATMs

Abstract

The special area of real-time objects moving according to monitoring methods is one of the soughtafter reasons for CNNs. This review work is pushed into the field of shooting firearms and pistols in areas covered by cameras. Domestic fire incidents, current impacts and fast-spreading fires are major problems with negative environmental impacts. Gun cruelty and mass shootings are also on the rise in obvious places on the planet. Such events are time-consuming and can be a monstrous test of life and property. The proposed work has now generated a meaningful learning model that enables the YOLOv3 computation to process video images by tuning to rationallyperceive those peculiarities and alert the experts. related. The latest model has an approval deficit of 0.2864, with a rendering rate of 45 edges per second, and has been compared with datasets such as IMFDB, UGR, and Fire Net with an accuracy of 90.3%, 83.6% and 85.5% independently.