Bean Leaf diseases encompass various ailments that can impact a bean plant’s leaves, notably angular leaf spots and bean rust, which can lead to poorer plant health and a reduced plant yield. Our study would benefit farmers from accurate and early bean disease detection since it enables them to respond promptly and effectively to manage and contain the disease's spread. For this purpose, this study utilizes Transfer Learning Models and compares them with another five proposed scratch Convolutional Neural Networks (CNN) models. The models were applied to an extensive dataset containing 1167 images of Bean Leaves divided into 3 classes namely Healthy, Bean Rust, and Angular Leaf Spot. After the entire evaluation, it was determined that the proposed CNN models performed better than the Transfer Learning Models, with the CNN-4 model having the highest accuracy of 97.68% and was hence concluded to be the best-fit classifier.