COMP 432 Machine Learning Project

This project represents the culmination of our work in COMP-432. Our objective was to develop a convolutional neural network capable of distinguishing between appropriate and inappropriate images. To accomplish this, we tested two distinct models employing different activation functions: softmax and sigmoid. Our results demonstrated that the model utilizing the softmax activation function outperformed the sigmoid model. Specifically, it achieved an impressive precision of 83% and an accuracy of 87% in identifying appropriate images. In the case of inappropriate images, the softmax model proved even more effective, with a precision of 91% and an accuracy of 87%.