Classification of Land Patterns from Indian Pines Dataset Using ResCNN

By utilising ResCNN, we were able to successfully classify the topographic attributes that were contained inside the Indian Pines dataset. The research also demonstrated how difficult it is to design the optimal configuration for a CNN, and it offered some recommendations for the parameter settings based on the findings of a grid search experiment. The recommendations were based on the findings of the experiment. The effect of patch sizes retrieved from images, the requirement of performing data augmentation, and the utilisation of regularisation techniques such as dropout and batch normalisation to avoid overfitting issues were also investigated in relation to the performance of the network. This was done to better understand how each of these factors affects the overall performance of the network. The primary finding of this work, which is that the convolutional neural network performed very well on minority classes in comparison to other approaches, is of the utmost importance.