Independent Researcher, USA.
Research Article
World Journal of Advanced Research and Reviews, 2022, 13(01), 875-884
Article DOI: 10.30574/wjarr.2022.13.1.0055
Received on 28 November 2021; revised on 04 January 2022; accepted on 06 January 2022
This research discusses the difference between Fully Connected (FC) and Dense layers during neural network image processing operations. FC layers from older deep learning models establish connections between each neuron and all neurons in the prior stage but generate high computational costs and a tendency to overfit the data. The specialized FC layer design, Dense layers, reduces parameter usage to improve processing efficiency in large-scale image data operations. This research investigates how the two different layer components affect the performance metrics used for image classification systems. This research reveals Dense layers achieve quicker training time and better efficiency than general FC layers in model operations. Typically, FC layers deliver superior results to Dense layers in extensive processing tasks that require critical model expression capabilities. The analysis demonstrates how optimization between processing speed and accuracy works for image processing deep learning models, which benefits professional developers in this field.
Fully Connected; Dense Layers; Image Processing; Model Performance; Computational Efficiency; Real-Time Applications
Preview Article PDF
Kolawole Favour Oluwafisayomi and James Andrew. Fully connected layer vs dense layer in image processing. World Journal of Advanced Research and Reviews, 2022, 13(1), 875-884. Article DOI: https://doi.org/10.30574/wjarr.2022.13.1.0055