| Proj No. | A3252-251 |
| Title | Performance Analysis of Deep Learning for Binary Semantic Segmentation |
| Summary | Objective: The goal of this project is to assess how many labeled scenes are required to achieve reliable performance in a binary semantic segmentation task using satellite SAR data. Specifically, the project will evaluate the relationship between the number of labeled samples and segmentation performance, using F1 score as the main evaluation metric. Data: The project will utilize publicly available Synthetic Aperture Radar (SAR) data from the European Space Agency’s Sentinel-1 mission. This data is free and open-access, making it suitable for scalable and reproducible analysis. Methodology: The student will implement a series of experiments to determine how the number of labeled scenes affects model performance. The task will involve binary classification of surface types (e.g., water vs. land) using semantic segmentation techniques. By systematically varying the number of labeled scenes and measuring the resulting F1 scores, the student will help establish practical guidelines for label efficiency in SAR-based segmentation tasks. |
| Supervisor | A/P Sang-Ho Yun (Loc:N2, Ext: ?) |
| Co-Supervisor | - |
| RI Co-Supervisor | - |
| Lab | Garage@EEE (for DIP only) (Loc: S1-B3c-26) |
| Single/Group: | Single |
| Area: | Digital Media Processing and Computer Engineering |
| ISP/RI/SMP/SCP?: |