A5059542928- Advanced Neural Network Applications
- Infrastructure Maintenance and Monitoring
- Domain Adaptation and Few-Shot Learning
- Vehicle License Plate Recognition
- Automated Road and Building Extraction
- Multimodal Machine Learning Applications
- Advanced Image Processing Techniques
- Industrial Vision Systems and Defect Detection
- Advanced Image and Video Retrieval Techniques
- Advanced Vision and Imaging
Sun Yat-sen University
2018-2020
Deep convolutional neural networks have been applied by automobile industries, Internet giants, and academic institutes to boost autonomous driving technologies; while progress has witnessed in environmental perception tasks, such as object detection driver state recognition, the scene-centric understanding identification still remain a virgin land. This mainly encompasses two key issues: 1) lack of shared large datasets with comprehensively annotated road scene information 2) difficulty...
Recently, research works have attempted the joint prediction of scene semantics and optical flow estimation, which demonstrate mutual improvement between both tasks. Besides, depth information is also indispensable for understanding, disparity estimation necessary outputting dense maps. Such task shares a great similarity with since they can all be cast into problem capturing difference at location two image frames. However, as far we know, currently there are few networks learning semantic...
Recently, the focus of semantic segmentation research has shifted to aggregation context prior and refined boundary. A typical network adopts modules extract rich features. It also utilizes top-down connection skips connections for refining boundary details. But it still remains disadvantage, an obvious fact is that problem false occurs as object very different textures. The fusion weak low-level features leads degradation. To tackle issue, we propose a simple yet effective network, which...
Deep residual network and multi-scale context module become two key ingredients in recent researches which made great progress semantic segmentation tasks. The features utilized by these networks are also proven to be effective refining object boundaries. However, regular convolution kernels inherently limited geometric transformation due the fixed structure. Additionally, when integrating information boundary detection, capability of global attention starts degrade problem inconsistent...
Road scene segmentation is of great significance in intelligent transportation system for different applications such as autonomous driving and semantic map building. Despite progress this field with the deep learning methods, there are still many difficulties robust small objects same type sizes scenes. In paper, we propose a new pyramid architecture segmentation, which top-down lateral connections multi-scale feature maps building, sufficiently incorporate momentous global scenery prior....