Volume 18, No. 4, 2021
Classification Of Hyperspectral Images Using Densenet-264 With Tensorflow
G. Narendra , Dr. D. Siva kumar
Abstract
The DenseNet-264 uses an additive principle to merge the previous layer with the next layer and further it concatenates the previous layer output with next layer. This DenseNet-264 model improves the declined accuracy produced due to vanishing gradients in case of high-level neural network. Hence, the information is preserved in longer paths between the source layer and destination layer without getting vanished in between the layers. In this paper, we study the DenseNet-264 top-5 metric to classify the Hyper-Spectral Image (HSI) features from the input HSI. The simulation is conducted on several HSI images to test the efficacy of the model against various datasets, and the results of simulation shows that the proposed method achieves 99% accuracy than the other existing classifier.
Pages: 157-170
Keywords: Classification, Hyperspectral Image, DenseNet-264, Tensorflow