Data visualization through Spark/pyspark after applying PCA and Classification on hyperspectral images

You'll be working with hyperspectral satellite images containing 242 bands. So the dataset will be divided into three tilesets 1) cloudy images, LULC(land use and land cover) images which include water, land, urban areas and forest, Crops which includes almost 7-8 variety of crops. We have to use unsupervised techniques like SVD, NMF, or KNN, etc. to segment the hyperspectral imagery into classes such as land, water, etc using band information from all the 242 bands of hyperspectral data. You can write your code on pyspark / spark to ensure that the time required for running the PCA and knn/ clustering reduces using a spark cluster at the backend. you can also record and compare the time taken for running the model on a simple ec2 instance vs on a spark cluster. You have to find the optimal number of distinct classes in every tileset / AoI that you select. you could scale up the modeling pipeline task by breaking the tileset in windows and use parallel processing technology for the same.