Bifa Liang

ORCID: 0000-0003-2200-9228
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Advanced Vision and Imaging
  • Advanced Image Processing Techniques
  • Image Processing Techniques and Applications
  • Optical Coherence Tomography Applications
  • Advanced Image and Video Retrieval Techniques
  • Autonomous Vehicle Technology and Safety
  • Advanced Neural Network Applications
  • Visual Attention and Saliency Detection
  • Advanced Fluorescence Microscopy Techniques
  • Image Enhancement Techniques

Guangzhou University
2022-2023

Object detection in point clouds is a critical component most autonomous driving systems. In this paper, order to improve the effectiveness of image feature extraction and accuracy clouds, pillar-based 3D cloud object algorithm with multiattention mechanism proposed, which includes three attention mechanisms SOCA, SOPA, SAPI. The results show that recognition optimized for cars, pedestrians, cyclists on KITTI dataset significantly improved benchmarks BEV 3D. Despite using only LiDAR, our...

10.1155/2023/5603123 article EN cc-by Wireless Communications and Mobile Computing 2023-02-09

Sensing and communications are dispensable for autonomous vehicles IoT. One key task in driving is the sensing of 3D information surrounding a vehicle. Most existing stereo disparity prediction networks pursue accurate maps on high-performance GPUs with high energy consumption. While very few achieving real-time resource-constrained edge devices hardly satisfactory terms accuracy. To tackle this, we propose lightweight efficient three-stage network named HRSNet matching energy-efficient...

10.1109/tgcn.2023.3233963 article EN IEEE Transactions on Green Communications and Networking 2023-01-03

3D perception is an essential capability of autonomous vehicles. Most state-of-the-art stereo matching networks pursue higher prediction accuracy at the cost inference speed. However, high demand on computational resource pushes hardware, hindering practical applications matching. In this paper, we propose RMCNet, a novel re-parameterized coarse-to-fine network for RMCNet achieves real time edge devices while outputs accurate disparity maps. To reduce computing complexity convolution in...

10.1109/tits.2023.3295930 article EN IEEE Transactions on Intelligent Transportation Systems 2023-07-25

Stereo matching is a classical problem in computer vision. It has been widely used many fields, especially autonomous driving recent years. Two key aspects of speed and accuracy are both desirable but conflicting characteristics driving. In this paper, we present CMNet, lightweight stereo architecture for improving the trade-off between on resource-limited devices. A novel feature extraction network consisted patch embedding layer ConvMLP-mixer proposed. The enhances receptive field makes...

10.1109/tvt.2022.3206612 article EN IEEE Transactions on Vehicular Technology 2022-09-14

Depth estimation is an essential element to constitute 3D perceiving ability of autonomous vehicles. Real-time inference on power- or memory-constrained devices would expedite the progress driving. In this paper, a lightweight stereo matching network proposed simultaneously achieve high accuracy and fast time. A novel feature extractor using patch embeddings for downsampling together with peculiar pyramidal strategy obtain accurate disparity maps less computational resources. By constructing...

10.1109/tvt.2023.3284011 article EN IEEE Transactions on Vehicular Technology 2023-01-01

Stereo matching is an important component technology that constitutes the 3D perception capability of autonomous vehicles. On resource-constrained edge devices, it very to compute in real-time with low time. However, most stereo networks focus on generating disparity maps high-end GPUs, which do not meet requirements devices. To solve this problem, a new network proposed paper achieve The greatly improves inference speed by constructing low-resolution feature extractor, and using multi-stage...

10.1109/access.2023.3297052 article EN cc-by-nc-nd IEEE Access 2023-01-01
Coming Soon ...