Osteoarthritis is the most prevalent form of arthritis, Millions of individuals around the globe are impacted by Osteoarthritis, which is the most prevalent type of arthritis. While it can harm any joint in the body, this condition is frequently observed in joints such as the hands, knees, hips, and spine. Generally, OA is prevalent among the elderly, while one of the areas that are frequently affected by Osteoarthritis is the knee. Knee Osteoarthritis happens when your knee joint cartilage wears out or is damaged. Pain is the most common symptom of osteoarthritis in the knee. Knee OA is complicated to detect, diagnose and treat. There is a demand for early and automatic detection of Knee OA. The objective of this study is to develop an application for the early prediction of knee osteoarthritis severity using deep learning approaches along with various neuron-wise and layer-wise visualization methods and feature engineering techniques using convolutional neural network (CNN) architectures. Knee images dataset is used to train the model, enabling it to identify and quantify the severity of osteoarthritis. The results show promising potential for the application of this technology in clinical settings, where early intervention can improve patient outcomes and reduce healthcare costs. The proposed algorithm trains and tests three architectures—Manual network, VGG-16 and Lenet-5.