Analyzing Baseline Models for Optimizing Deep Neural Networks in Resource-Constrained Environments

In order to identify a potential baseline model for optimizing deep neural network (DNN) models for deployment in resource-constrained environments (RCE), this research presents comprehensive and empirical analyses of advanced DNN models. While DNNs perform exceptionally well in various applications, deploying them on RCE devices, such as wearables and mobile phones, is challenging. The article investigates several popular optimization models, including MobileNetV۲, ResNet۵۰, InceptionV۳, DenseNet۱۲۱, and EfficientNetB۱. Each of these models is subjected to a comprehensive and empirical analysis, considering their unique characteristics, usability, efficiency, strengths, and weaknesses in the context of efficient deployment in RCE scenarios. The analysis aims to uncover how each model can be optimized to function effectively within the limited computational and memory resources typical of RCE devices. By examining these models, the research identifies critical factors that influence their performance on RCE devices, such as model size, computational complexity, and inference speed. The findings highlight the trade-offs between model accuracy and resource efficiency, providing insights into how DNN models can be tailored to meet the constraints of mobile and wearable technology. The review concludes by recommending EfficientNetB۱ as a baseline model for future research aimed at developing efficiency-focused DNN models for image classification on RCE devices. This recommendation is based on EfficientNetB۱'s balance of performance and efficiency, making it a suitable starting point for further optimization in resource-constrained settings.

کلیدواژه ها:

Baseline model ، Deep Neural Network ، Image classification ، optimization ، Resource-constrained environments