Abstract: The deployment of Convolutional Neural Networks (CNNs) on resource-constrained edge devices for inference is challenging due to its computation, memory, energy, and bandwidth requirements.