Performance Comparison of Convolution Neural Networks on Weather Dataset
Aravind S Raj, Anju S. Pillai · 2022 4th International Conference on Inventive Research in Computing Applications (ICIRCA) · 2022
Deep neural networks (DNNs) are quickly becoming a crucial enabling technology in a wide range of application fields. However, due to the long inference time and resource requirement, it is almost infeasible to implement DNNs on battery-powered resource-constrained embedded systems. Simply transferring computational load to the cloud is frequently discouraged due to security concerns, excessive delay, or a lack of connection. In the proposed work, four notable convolutional neural network-based models explicitly, VGG19, MobileNet V2, InceptionResNet V2, and Inception V3 have been considered. The proficiency of the considered models mentioned has been assessed on the custom weather dataset for multiclass classification. CNNs are genuinely strong for image characterization as the possibility of dimensionality decline suits the gigantic number of parameters in a picture.