Knee-related disorders affect individuals across all age groups, with osteoarthritis (OA) recognized as one of the most common and debilitating conditions. A key radiographic indicator of OA progression is the narrowing of the joint space, which reflects underlying cartilage degradation. Traditionally, this evaluation is performed manually by expert radiologists through the visual inspection of knee X-ray images—an approach that is time-consuming, labor-intensive, and prone to subjectivity. In this study, we propose an automated, computer vision-based diagnostic system to assist radiologists in the detection and classification of knee osteoarthritis. Our method employs deep convolutional neural network architectures—specifically ResNet and DenseNet—for robust feature extraction from preprocessed and augmented knee X-ray images. After extensive experimentation, DenseNet emerged as the more effective model, providing superior performance in distinguishing between normal and osteoarthritic joints. The classification is performed with reference to standard knee joint space width metrics and is validated using a substantial Kellgren-Lawrence (KL) graded dataset. The proposed model achieved a detection accuracy of 97.36%, surpassing several existing approaches and demonstrating its potential as a reliable clinical decision support tool for early and accurate OA diagnosis.