STM Journals Peer-reviewed research from Consortium e-Learning Network

Journal of Image Processing & Pattern Recognition Progress Review Article

Energy-efficient Image Classification on Edge Devices: Implementation and Evaluation

  1. Shital Thakkar Department of Electronics and Communication Engineering, Dharmsinh Desai University (DDU), Nadiad
  2. Arpit Acharya Department of Electronics and Communication Engineering, Dharmsinh Desai University (DDU), Nadiad
  3. Richa Singh Department of Electronics and Communication Engineering, Dharmsinh Desai University (DDU), Nadiad
  4. Harsh Bhalala Department of Electronics and Communication Engineering, Dharmsinh Desai University (DDU), Nadiad
  5. Pinkesh Patel Department of Electronics and Communication Engineering, Dharmsinh Desai University (DDU), Nadiad
  6. Vinay Thumar Department of Electronics and Communication Engineering, Dharmsinh Desai University (DDU), Nadiad

Abstract

Image classification is a computer vision problem where an algorithm determines a class or label for a given image. Various real-time applications like object recognition, medical diagnosis, person recognition, etc. Image classification property on edge devices is useful for autonomous vehicles, surveillance, and healthcare and internet of things deployments. The advancement of deep learning based methods and graphics processing units (GPU) devices allows efficient processing locally. The study utilizes a GPU-powered board, recognized for its strength in visual computing, to run a convolutional neural network (CNN) classification algorithm. Deep learning with GPUs speeds up image analysis by extracting features and constructing neural networks, enabling efficient processing of both static and real-time data. The combination of edge device efficiency and GPU acceleration illustrates the potential for strong image classification in practical scenarios. We demonstrated classification tasks such as binary classification of cats versus dogs and multi-class classification on the CIFAR10 benchmark dataset.

Keywords

References (11)

  1. Krizhevsky A, Sutskever I, Hinton GE. ImageNet classification with deep convolutional neural networks. Communications of the ACM. 2017;60(6):84-90. doi:10.1145/3065386
  2. Aliady H, Utari DT. GPU based image classification using convolutional neural network chicken dishes classification. Int J Adv Soft Comput Appl. 2018;10:1–3.
  3. Ang LM, Seng KP. GPU-Based Embedded Intelligence Architectures and Applications. Electronics. 2021;10(8):952. doi:10.3390/electronics10080952
  4. Simonyan K, Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition. [Preprint]. ArXiv:1409.1556. 2014 Sept 4.
  5. Gavali P, Banu JS. Deep Convolutional Neural Network for Image Classification on CUDA Platform. Deep Learning and Parallel Computing Environment for Bioengineering Systems. 2019:99-122. doi:10.1016/b978-0-12-816718-2.00013-0
  6. (2024). NVIDIA DGX Platform. [online] Available from: https://www.nvidia.com/en-us/data-center/dgx-platform/
  7. Bhavnagri K. (2019). Convolutional Neural Networks: Basic Theory in a Nutshell. [online] The Data Science Swiss Army Knife. Available from: https://www.kamwithk.com/convolutional-neural-networks-basic-theory-in-a-nutshell
  8. Huang G, Liu Z, Van Der Maaten L, Weinberger KQ. Densely Connected Convolutional Networks. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2017:2261-2269. doi:10.1109/cvpr.2017.243
  9. Tan M. Efficientnet: Rethinking Model Scaling for Convolutional Neural Networks. [Preprint]. ArXiv:1905.11946. 2019.
  10. Davidson RL, Bridges CP. Error Resilient GPU Accelerated Image Processing for Space Applications. IEEE Transactions on Parallel and Distributed Systems. 2018;29(9):1990-2003. doi:10.1109/tpds.2018.2812853
  11. Krizhevsky A, Nair V, Hinton G. (2024). CIFAR-10 and CIFAR-100 datasets. [online] Toronto (Canada): University of Toronto. Available from: https://www.cs.toronto.edu/~kriz/cifar.html
Support