Knee osteoarthritis, prevalent among the elderly, involves gradual articular cartilage degradation. The Kellgren-Lawrence grading system assesses severity using radiographic features. Our study incorporates EfficientNetB5, offering insights into its potential efficacy for OA classification. Consolidating Kellgren-Lawrence grades significantly improves accuracy. Algorithm scores align with MOST dataset and radiologist .The Method Used is knee osteoarthritis (OA) classification study, we employed the advanced EfficientNetB5 architecture. The methodology included meticulous dataset curation, transfer learning, and strategic data augmentation to enhance model performance, ensuring its adaptability to varying severity levels. The results demonstrates commendable proficiency with an accuracy of 64%, showcasing its effectiveness in accurately classifying different severity levels of knee OA. This underscores the potential of EfficientNetB5 in precise knee osteoarthritis severity assessment.