This chapter will delve into the growing challenge of deploying complex neural networks in resource-constrained environments, particularly within the realm of image and video coding. As neural networks become more sophisticated, their demand for computational resources escalates, complicating their application in edge and mobile devices. To address this, lightweight techniques like pruning, quantization, and knowledge distillation are used to decrease model size and computation complexity without sacrificing performance. By streamlining neural networks, these methods facilitate more efficient deployments and enhance real-time processing capabilities. This chapter will focus on various lightweight approaches and hardware implementations specifically designed for learning-based image and video compression networks.