| Proj No. | A1066-251 |
| Title | Category-Aware Filtering for Robust Object Detection in Autonomous Driving |
| Summary | The category-aware filtering solution presents an innovative approach to improving object detection accuracy in autonomous driving by integrating feature consistency analysis and post-processing refinement. Existing object detectors, such as YOLO and Faster R-CNN, often misclassify objects due to overlapping visual features or biased training data. For example, poles or signposts near a bus stop may be misclassified as pedestrians due to similar shape and texture patterns. A significant challenge involves identifying and correcting such misclassifications in real time without compromising the detection system’s efficiency. Students will investigate a Bayesian classifier and similarity-based correction framework for real-time category-aware filtering in complex driving environments. The proposed method will utilize real-world traffic datasets and a high-fidelity simulation environment (e.g., CARLA or KITTI) to model category misclassification scenarios. The research will focus on improving the consistency between predicted object class and extracted visual features (e.g., shape, texture, motion). The solution will involve feature extraction, Bayesian classification, and similarity-based post-processing to correct potential classification errors. The ultimate objective is to advance the scientific knowledge in real-time object detection and classification under complex urban driving scenarios. The student candidates should possess the following qualities: 1. A strong grasp of mathematical concepts (e.g., Bayesian inference, feature similarity metrics) and the ability to review scholarly papers. 2. Previous experience with deep learning frameworks (e.g., PyTorch, TensorFlow). 3. A commitment to dedicate time to experimental work. |
| Supervisor | Prof Su Rong (Loc:S1 > S1 B1B > S1 B1B 59, Ext: +65 67906042) |
| Co-Supervisor | - |
| RI Co-Supervisor | - |
| Lab | Centre for Advanced Robotics Technology Innovation (CARTIN) (Loc: S2.1-B3-01) |
| Single/Group: | Single |
| Area: | Intelligent Systems and Control Engineering |
| ISP/RI/SMP/SCP?: |