Baseline model for deep neural networks in resource-constrained environments: an empirical investigation

Abstract

This paper presents an empirical study on advanced Deep Neural Network (DNN) models, with a focus on identifying potential baseline models for efficient deployment in resource-constrained environments (RCE). The systematic evaluation encompasses ten state-of-the-art pre-trained DNN models: ResNet50, InceptionResNetV2, InceptionV3, MobileNet, MobileNetV2, EfficientNetB0, EfficientNetB1, EfficientNetB2, DenseNet121, and Xception, within the context of an RCE setting. Evaluation criteria, such as parameters (indicating model complexity), storage space (reflecting storage requirements), CPU usage time (for realtime applications), and accuracy (reflecting prediction truth), are considered through systematic experimental procedures. The results highlight MobileNet's excellent trade-off between accuracy and resource requirements, especially in terms of CPU and storage consumption, in experimental scenarios where image predictions are performed on an RCE device. Consequently, MobileNet emerges as a suitable baseline model for future DNNs developed specifically for RCE image classification. The study's conclusions endorse MobileNet as a baseline model for transfer learning techniques (used in DNN design), providing valuable insights for optimizing DNN models in resource-constrained scenarios. This approach enhances the creation of efficiency-focused and lightweight DNN models, improving their application and efficacy in resource-constrained environments. Future research will leverage the identified MobileNet model as a foundation to create a new DNN model tailored for efficiency-driven image classification applications in RCE devices.