Optimizing Image Classification with Inverse Depthwise Separable Convolution for Edge Devices

Deep Neural Networks (DNNs) have catalyzed a revolution in the field of object classification, achieving levels of accuracy that occasionally surpass human capabilities. Despite these significant advancements, the deployment of DNNs on edge devices presents new challenges due to inherent constraints in size, computational resources, and energy efficiency. Addressing these challenges necessitates innovative solutions that strike a balance between computational efficiency and classification accuracy. In this research paper, we introduce a novel Image Classification (IC) model that leverages Inverse Depthwise Separable Convolution, a technique specifically designed to optimize both accuracy and computational cost. With a lean structure of only 0.20M parameters, our proposed model achieves an impressive accuracy of 90.1% on the CIFAR-10 dataset. Moreover, the model’s compact size of 2.9 MB makes it an ideal candidate for deployment on edge/hardware devices. This work represents a significant contribution to the ongoing efforts to develop lightweight and efficient models without compromising performance, thereby facilitating broader applications in real-world edge computing scenarios.