Design of Efficient Algorithms for Video Surveillance Applications using Artificial Intelligence

10/05/2022

∙

by   Mohana, et al.

∙

0

∙

share

Object detection and tracking algorithms such as YOLO(You Look Only Once Version V1 to V3), SSD and SORT implemented on COCO and indigenous data set for traffic surveillance and evaluated using the performance metrics such as True Positive (TP), True Negative (TN), False Positive (FP), False Negative (FN), mean Average Precession (mAP). The designed CNN trained on dataset (small and large) had similar performance on test dataset, however the CNN trained on the large datasets that had larger intra-class variations was able classify a greater number of vehicles belonging to light and two-wheeler class. It achieved a validation accuracy of 98%. VGG16 achieved an accuracy of 97% followed by MobileNetV2 and InceptionV3 with 75% and 50% accuracy respectively