AutoMap is a utility tool designed to automate road infrastructure detection using satellite imagery. By leveraging deep learning models like UNet and Attention UNet, it processes satellite images providing accurate predictions for large-scale mapping.
This project was developed as a Research Project as a part of KAIST's (Korea Advanced Institute of Technology) summer research program.
Selected among 18 students worldwide for the prestigious summer research program at KAIST, focusing on deep learning research
Developed UNet and Attention UNet models to automate road infrastructure detection using satellite imagery, aimed at reducing manual labor in road mapping.
Achieved a strong 0.73 F1 score on the final model.
Curated and preprocessed data from datasets like Massachusetts and SpaceNet, correcting known errors for model enhancement.