Autism spectrum disorder, more commonly referred to as Autism, is a neurological developmental condition which impacts social interaction, communication, behavior regulation and sensory processing. It is usually observed at the early age in children. In proposed work, an autism detection method using the eye-tracking images is presented. The classification of autistic and non-autistic patients is performed using deep learning methods—GoogleNet and ResNet-18. Additionally, hybrid method by combining GoogleNet and ResNet-18 with Support Vector Machine (SVM) is also analyzed for ASD prediction. The proposed model achieved an accuracy of 99% using ResNet-18 model with eye-tracking images as input for patients. The findings illustrate the potential of proposed model in facilitating early intervention and improving outcomes for affected individuals.