Disguised Face Detection using Machine Learning

The crime rate has been increasing day by day. According to India today, the police have said that 3,06,389 crime cases were reported in 2021 alone. To reduce the crime rate, faster action and stricter law policies are required. Due to the increasing crime rate, it becomes a very difficult task for the police to catch the criminals and address the issue quickly. One strong piece of evidence that helps the Police identify criminals is the CCTV footage. Unfortunately, the images in CCTV footage are generally blurred. Also, when criminals disguise themselves and commit a crime, it becomes even more difficult in identifying their faces. Hence, more time is required to resolve cases. To solve this problem, a tool was developed for disguised detection of images using machine learning algorithms. This tool includes age and gender detection of disguised images, an image search engine that can retrieve similar images, and the calculation of percentage similarity between the images. In our paper testing is performed using various algorithms like CNN, KNN, SVM, Logistic Regression and Decision tree [1]. Apart from this, a dataset is also generated with real time images, and further applied various noises, blurred images, and tested the model. We have observed that using SVM, an accuracy of around 81% is achieved and CNN has performed extremely well with an accuracy of 90.6%.