Project Description

Every year, a large amount of the population reconciles gun-related violence all over the world.A computer-based fully automated system to identify basic armaments, particularly knives, handguns and rifles was necessary. Applying this model to a surveillance system, we can attempt to save human life and accomplish a reduction in the rate of crime.The dataset available for this problem was scares and manual annotation would be time consuming, hence we generated our own data set. Generated 20,000 image dataset of 4 scenarios with characters and weapons using Python multithreading for parallelization. Trained and tested the generated image dataset using ResNet 50, Inception v3, MobileNet by splitting the dataset as necessary. Compared the accuracy and training time of all the three weapon detection algorithms.