Hand gesture recognition is a burgeoning area of study with wide-ranging applications, including human–computer interaction, sign language translation, and virtual reality control. In this work, we introduce an innovative lightweight DL model designed for robust HGR and provide an in-depth analysis of its classification performance. We meticulously evaluate key performance metrics such as precision, detection rate, Jaccard index, and false positive rate and compare these results with existing methodologies. Our proposed model stands out by surpassing the performance of current models. It exhibits higher accuracy rates during both training and validation, a substantially improved detection rate, and an almost perfect Jaccard index. Most impressively, our model achieves a remarkable false positive rate of zero, showcasing its exceptional reliability in distinguishing between positive and negative instances. These findings underscore the effectiveness of our approach in addressing challenges related to overlapping gestures, unbalanced class distributions, and precise HGR. In our conclusion, we identify areas for potential enhancement and further research. To clarify the novelty and research gap of our proposed model, we emphasise its lightweight nature and articulate how it differs from existing methods. This distinction lies in its capacity to deliver accurate and efficient HGR while imposing minimal computational burden, setting it apart from more resource-intensive approaches. In summary, our study and potential to revolutionise practical applications that demand dependable and effective hand gesture recognition systems, bridging existing gaps in the area.