Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{185202,
        author = {Prof. Ayesha Asif Sayyad and Vivendra Kumar and Yashraj Vhalgade},
        title = {Optimizing Neural Networks for Deployment on IoT and Mobile Devices using Deep learning},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {5},
        pages = {601-613},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=185202},
        abstract = {The rapid growth of the Internet of Things (IoT) and mobile computing has created a rising need to run intelligent deep learning models directly on small, resource-limited devices. Unfortunately, traditional neural networks are often too heavy, demanding large amounts of memory, storage, and power—far beyond what lightweight embedded platforms can handle. This research looks at how to make these models more efficient and practical for IoT and mobile devices. It explores techniques like pruning unnecessary network connections, compressing weights through quantization, transferring knowledge from larger models to smaller ones, and designing lighter architectures such as MobileNet and TinyML frameworks. Together, these methods aim to shrink model size, speed up inference, and cut down energy use—all while maintaining strong accuracy. The study also discusses the balance between efficiency and accuracy, with real-world examples in areas like health monitoring, smart home systems, and industrial IoT. It further examines how different optimization strategies affect energy consumption on edge devices. The results show that with the right optimization, deep learning models can come close to state-of-the-art performance while being lightweight enough for real-time use on low-power devices. This opens the door to scalable, intelligent edge computing that can power the next generation of smart, connected systems},
        keywords = {The goal of this work is to make deep learning sufficiently efficient to operate directly on tiny, low-power devices (wearables, smartphones, and Internet of Things devices). It reduces neural networks with little loss of accuracy by using methods like knowledge distillation, quantization, pruning, and model compression. In conjunction with lightweight architectures (e.g., MobileNet, TinyML), the objective is to lower energy consumption and enable real-time inference on the device (edge AI), enabling smart applications in industrial IoT, smart homes, health, etc.},
        month = {October},
        }