Autism Spectrum Disorder (ASD) is a neurological and developmental disorder that affects people’s perceptions and interactions with the world. It is a spectrum disorder which means it affects the individuals in a wide and different range of ways. Traditionally ASD is detected via behaviour assessments and tests which can be subjective and resulting in delayed results and inconsistencies. As these diagnostic methods are challenging, early detection remains difficult. This paper proposes a deep learning based ASD detection method using Structural MRI (sMRI), initially the sMRI slice are enhanced using an improved nimble filter and Region of Interest (ROI) is extracted using FSL Brain Extraction Tool (BET). Finally the classification is performed using EfficientNet-B5 model with a weight balancing. The system is evaluated using a 5-fold cross validation approach to ensure generalizability and robustness. The experimentation results showcase high performance in terms of evaluation metrics, indicating the effectiveness of the proposed system, thereby providing an effective approach for detecting ASD.