This project introduces a novel hand gesture recognition system utilizing transfer learning and the incorporation of sophisticated classification methodologies. The system uses pre-trained models, like MobileNetV2 and ResNetV2, to get high-quality features. We made these architectures better at classifying hand gestures by fine-tuning them. MobileNetV2 was the best choice for this job because it was more accurate and faster at computing. To improve performance, MediaPipe was added to find hand landmarks and make it easier for both a keypoint classifier and a point history classifier to work. These parts work together well to improve gesture recognition by recording changes over time and in space based on landmarks. The last model works very well and can recognize gestures in real time, showing that it could be useful for things like sign language interpretation and human-computer interaction.