A Comprehensive Analysis of Interaction Based Segmentation Using Deep Learning Techniques

Image segmentation, from a human's point of view, is the separation of an image into distinct, informative sections that do not overlap. Segmenting an image is a common technique used in image analysis to isolate certain features of interest. When we are working on satellite images we found some problems with existing approaches. These are: The noise and the artifacts in the images are the major problems for segmentation processing especially in satellite images. The inability of the edge detection based algorithms to mask the whole segmented. Object detection and segmentation using automation major issue for low visibility. In this paper, we use CNN based U-net model rapid, robust, and more accurate generalized prediction from high-resolution satellite images. In this paper, we use a convolutional neural network and main deep learning model for semantic segmentation to evaluate the performance factors like precision, recall, FPR, TPR, IOU, MIoU and F score.