| Proj No. | A3145-251 |
| Title | 3D Long Range Object Detection Under Rainy Conditions |
| Summary | Rainy weather and unfavourable lighting conditions are infamous for severely reducing the quality of the visual data feed of autonomous vehicles (AV). As a result, camera-based object detectors become unreliable in complicated scenarios, leading to an unacceptable increase in missed or false positive detections. As such, it is foreseen that LiDAR-based object detection methods will be more effective than vision-based methods, because they are not susceptible to light conditions, and can detect and classify targets, providing rich 3D information. Despite this, the effectiveness of LiDAR data processing algorithms in rain remains largely unexplored. It is predicted that the reduced quality of LiDAR data decreases the reliability of 3D long-range object detection. In this project, the student is tasked with developing a 3D LiDAR object detector for long-range object detection to assess the performance under different rain levels. Popular frameworks such as Centerpoint may be considered as the basis for the 3D object detector. Common metrics for object detection such as Average Precision, Average Recall, and Confusion Matrix will be adopted to measure the performance. The student would be interfacing with simulated data or with data recorded at A*STAR (which the student is expected to annotate) for evaluating their approach and can expect to develop a good understanding of 3D object detection. As an optional goal, using transformers or other methods to improve the deep learning detection accuracy may be explored. |
| Supervisor | A/P Soong Boon Hee (Loc:S2 > S2 B2C > S2 B2C 115, Ext: +65 67905398) |
| Co-Supervisor | - |
| RI Co-Supervisor | - |
| Lab | Centre for Information Sciences & System (CISS) (Loc: S2-B4b-05) |
| Single/Group: | Single |
| Area: | Wireless and Communications Engineering |
| ISP/RI/SMP/SCP?: |