Real Time Theft Detection Using YOLOv5 Object Detection Model

Deep learning is a machine learning approach that trains computers to accomplish things that humans do instinctively. It teaches the compute/AI to visualize the world as humans see it and visualize it. This is done possible by mimicking the human brain along with its complex network of neurons. The concept of neural network is widely considered in such cases where the AI designs its own brain/neural network by taking in examples. While approximate predictions can be done with a single layer neural network, additional hidden layers can lead for higher accuracies and also better optimization. Object detection is a subset of deep learning, which involves the detection of objects within frames. Using the above technologies, the concept of the theft detection can be met. Theft of valuable objects is increasing day by day and is an extremely important problem with respect to the global scale and due to the fact that there are only manual systems present to detect theft, many of the thieves/robbers never get caught and repeat their crimes. Thus, the need for an automated system which detects theft with good accuracy and immediately alerts the owners must be developed. You Only Look Once (YOLO) v5 model, an object detection model that is best suited for our case regarding theft detection is considered here. The YOLO v5 object detection model is then trained using our own dataset. After its training its accuracies and prediction of both objects and theft is done in real time along with an alert being issued to concerned authorities and the owner of those objects/items.