1Mrs. B. NIRUPAMA, 2KONDRA ANJALI REDDY, 3KOPPELA PRANAVI, 4DONGALA MADHU 03 05 In recent years ensuring safety in publicenvironments has become a critical challenge dueto the rising number of weaponrelated incidentsTraditional surveillance systems such as CCTV relyheavily on human monitoring which is ofteninefficient due to fatigue distraction and delayedresponse This project presents an intelligent deeplearningbased system for dangerous objectdetection in public spaces to enhance security andreduce dependency on manual surveillance Theproposed system utilizes advanced object detectionalgorithms such as YOLO Faster RCNN and SSDMobileNet to identify weapons like guns andknives in realtime from video streams The systemis trained on a diverse dataset consisting of weaponand nonweapon images including confusionobjects to reduce false positives Imagepreprocessing techniques such as normalizationaugmentation and enhancement are applied toimprove detection performance in challengingconditions like low lighting and crowdedenvironments The model evaluates performanceusing precision recall F1score and mean averageprecision metrics ensuring accurate detectionUpon identifying a dangerous object the systemgenerates immediate alerts allowing authorities totake timely action This automated approachimproves surveillance efficiency reduces humanworkload and enhances response time The systemis scalable costeffective and suitable fordeployment in areas such as airports railwaystations malls and educational institutionsOverall the proposed solution contributessignificantly to public safety by providing areliable realtime and intelligent surveillancemechanism Copyright (c) Journal of Science Engineering Technology and Management Science. All rights reserved 2026