Papers
6
Total Citations
160
H-Index
4
About
Wei-Chen Chiu is a computer vision researcher whose work spans semantic scene understanding, depth estimation, and multimodal perception — areas that sit at the intersection of deep learning and practical robotics applications. He is perhaps best known for his development of STD2P (Spatio-Temporal Data-Driven Pooling), a superpixel-based multi-view convolutional neural network for RGBD semantic segmentation, which has garnered over 130 citations and stands as his most influential contribution to the field. This work demonstrated how temporal and spatial context from multiple views can be leveraged to significantly improve segmentation quality in indoor video environments. Beyond segmentation, Chiu has made notable contributions to depth sensing, exploring optimal filter learning for cross-modal stereo with projected patterns — work inspired by the widespread adoption of the Kinect sensor — as well as lightweight monocular depth estimation for real-time autonomous driving applications. His research also extends to 3D scene layout understanding, evidenced by the LayoutMP3D annotation dataset for panoramic environments, and unsupervised face recognition through disentanglement and self-augmentation, reflecting a broad interest in self-adapting vision systems. Collectively, his work addresses fundamental challenges in making machines perceive and interpret complex environments reliably and efficiently.
Research Focus
Key Achievements
Top Papers
- 1STD2P: RGBD Semantic Segmentation Using Spatio-Temporal Data-Driven Pooling132 citations · 2017
- 2
- 3LayoutMP3D: Layout Annotation of Matterport3D8 citations · 2020
- 4
- 5
- 6