Abstract: Detecting Buildings from high resolution satellite images is incredibly helpful in map making, land use analysis, urban planning etc.  If a human expert wants to manually extract this valuable data, it is difficult. So there is one potential solution to extract this data by using automatic techniques. In this project we are going to use deep learning approach to detect buildings from the satellite images. We have downloaded a large amount of dataset because deep learning model requires a huge dataset for training the model. Then pre-processing of these image like cropping, resizing and removal of noise by using Gaussian filter is being done. Split the dataset into training and validation set. After gathering all the dataset, we should train the model. In training process, the dataset undergoes through convolutional neural networks for different layers. Once training is completed the model has been built and we test the model by applying different input images and finally we get only the detected buildings as output. In this project we have also done the greenery detection by using HSV (Hue, Saturation, Value) colour format. We are using python for coding purpose.

Keywords: Satellite Images, Building Detection, Convolutional Neural Networks and Deep learning.

| DOI: 10.17148/IJIREEICE.2020.8904