Deep Learning-based Fine-tuned Convolutional Neural Network Model for Google Map Image Analysis

This research focuses on creating and assessing a Convolutional Neural Network (CNN) model for detecting different properties in Google Maps images. With the increased availability of geospatial data, the automated analysis of satellite and street-level images has become crucial for urban planning, environmental monitoring, and navigation systems. In the proposed work, the CNN model was trained on a dataset including 8000 annotated images, encompassing categories such as bridges, landmarks, industries, parks, playgrounds, and parking lots. The training approach includes preparing the images by resizing them to a consistent size, levelling pixel values, and enriching the dataset to promote variety. The CNN design comprised convolutional layers followed by pooling layers and fully connected layers for classification. The dataset was separated into training, validation, and test sets, with the model obtaining an accuracy of 97.48% on the test set at 20 epochs of training. Additionally, the loss reduced significantly across epochs, showing good learning and convergence of the model. The findings highlight the potential of the CNN model for reliably recognizing elements in Google Maps images, with implications for urban planning, transportation management, and environmental study. Further, the study might examine the incorporation of real-time data and other image attributes to increase the model's performance and applicability in varied scenarios.