Automatic Feature Extraction from Remote Sensing Images

The recent years have witnessed significant strides in deep learning technology, particularly in the fields of computer vision, natural language processing, and multimedia. This advancement has created new opportunities for overcoming the challenges associated with traditional methods of collecting geographic data. Traditional approaches often demand substantial resources in terms of manpower and materials, while also encountering obstacles such as object masking and variations in colors, widths, and shapes. To tackle these issues, the proposed project presents a novel approach that harnesses the power of deep learning, specifically the YOLOv5 algorithm based on Convolutional Neural Networks (CNNs), to automatically extract spectral-spatial features from high-resolution multi-spectral images. By utilizing YOLOv5, this method aims to revolutionize the field of geographic data collection by enhancing the effectiveness and efficiency of data extraction.