Pseudo Trained YOLO R_CNN Model for Weapon Detection with a Real-Time Kaggle Dataset

Abstract

The Recurrent Convolutional Neural Networks (RCNN) based deep learning models has been classified image patterns and deep features through layer architecture. In this world every country doesn’t encouraging violence, so that indirectly nations prohibiting usages of weapons to common people. This study proposes a novel YoLo Faster R-CNN based weapon detection algorithm for unusual weapon object detection. The proposed YoLo V3 R-CNN computer vision application can rapidly find weapons carried by people and highlighted through bounding-box-intimation. The work plan of this research is divided into two stages, at 1st stage pre-processing has been called to Faster R-CNN segmentation. The 2nd stage has been training the dataset as well as extracting 8-features (image_id, detection score, pixels-intensity, resolution, Aspect-ratio, PSNR, CC, SSIM) into.csv file. The labeling can be performed to RCNN-YoLo method such that getting real-time objects detection (Unusual things). The Confusion matrix has been generating performance measures in terms of accuracy 97.12%, SSIM 0.99, sensitivity 97.23%, and throughput 94.23% had been attained which are outperformance methodology.