In our current day firearms and other weaponry is becoming more and more common, being used for both lawful and unlawful ends. A wholly reactive response is unsatisfactory and so, prevention is key. And a major part of that prevention? Detection. There should be little doubt of a present threat.
Using a training set of 11,000 images, 25% of which are filtered through sustantial adversarial noise, we have trained a neural network utilizing the YOLOv5 object detection algorithm to spot and highlight knives and pistols with an average 81% accuracy.
Simply open the YOLOv5_Weapon_Detection.ipynb in jupyter notebooks or click the link below to access the file on Google colab.
