This repository contains the code for a road segmentation model using the U-Net architecture. The model is trained on the DeepGlobe Road Extraction Dataset and can be used to extract road networks from satellite images. Table of Contents
Introduction
Dataset
Model Architecture
Results
Road extraction from satellite imagery is an important task in the field of remote sensing and has various applications in urban planning, transportation management, and disaster response. This project aims to develop a deep learning model for accurate and efficient road segmentation.
The U-Net architecture, a widely used convolutional neural network for image segmentation, is employed in this project. The model is trained on the DeepGlobe Road Extraction Dataset, which contains satellite images and corresponding road masks. Dataset
The DeepGlobe Road Extraction Dataset is used for training and evaluating the road segmentation model. The dataset consists of 6,226 satellite images and their corresponding road masks, with a resolution of 1024x1024 pixels.
The model used in this project is a modified version of the U-Net architecture. It consists of a contracting path (encoder) and an expansive path (decoder), with skip connections between the corresponding layers of the encoder and decoder.
The main components of the model are:
Convolutional layers with ReLU activation
Max pooling layers
Batch normalization
Dropout layers
Upsampling layers
Concatenation of feature maps from the encoder and decoder
The final layer of the model uses a sigmoid activation function to produce a binary road segmentation mask.
The U-Net model trained on the DeepGlobe Road Extraction Dataset achieves a Jaccard coefficient (Intersection over Union) of 31.6 on the validation set.