Improving driver assistance system under adverse conditions using generative deep learning models | |
| Author | Win Win Phyo |
| Call Number | AIT Thesis no.DSAI-23-08 |
| Subject(s) | Driver assistance systems Deep learning (Machine learning) |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Data Science and Artificial Intelligence |
| Publisher | Asian Institute of Technology |
| Abstract | In smart driving systems, the identification of lane markings is foundational. How ever, consistently detecting these markings poses challenges due to degradation, oc clusion, illumination variations, light fluctuations, complex traffic scenarios, and the absence of lane lines. This research introduces a technique to pinpoint various lane markings, even under less-than-ideal conditions. The initial step involves the de tection and masking of lane boundaries using an object detection network. Subse quently, in situations where lane markings are obscured or absent, an image-to-image translation network helps regenerate them. This study employs the Yolact real-time instance segmentation model for object detection and the unpaired CycleGAN for image-to-image translation. In benchmark tests, the implemented approach recorded an impressive mAP of 97.01% on the TuSimple dataset, outshining other models like Faster R-CNN (which scored 78.90% on COCO) and MobileNetSSD (which achieved 72.40% on PASCALVOC). Additionally, using the Resnet50 backbone on a merged Bangkok and TuSimple dataset, an F1 score of 75.75%, which saw further enhancement post CycleGAN integration and image augmentations. The model’s speed efficacy was also assessed, achieving a commendable 91 FPS on the RTX 3080 Ti. |
| Year | 2023 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Data Science and Artificial Intelligence (DSAI) |
| Chairperson(s) | Dailey, Matthew N. |
| Examination Committee(s) | Chaklam Silpasuwanchai;Mongkol Ekpanyapong |
| Scholarship Donor(s) | Asian Institute of Technology |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2023 |