A5015741623- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- Robotics and Sensor-Based Localization
- Multimodal Machine Learning Applications
- 3D Shape Modeling and Analysis
- Remote Sensing and LiDAR Applications
- Domain Adaptation and Few-Shot Learning
- Integrated Energy Systems Optimization
- Topic Modeling
- Microgrid Control and Optimization
- Optimal Power Flow Distribution
Zhejiang University
2020-2021
In this paper, we propose a Monocular 3D Single Stage object Detector (M3DSSD) with feature alignment and asymmetric non-local attention. Current anchor-based monocular detection methods suffer from mismatching. To overcome this, two-step approach. the first step, shape is performed to enable receptive field of map focus on pre-defined anchors high confidence scores. second center used align features at 2D/3D centers. Further, it often difficult learn global information capture long-range...
The most recent 3D object detectors for point clouds rely on the coarse voxel-based representation rather than accurate point-based due to a higher box recall in Region Proposal Network (RPN). However, detection accuracy is severely restricted by information loss of pose details voxels. Different from considering cloud as voxel or only, we propose point-to-voxel feature learning approach voxelize with both point-wise semantic and local spatial features, which maintains voxel-wise features...
Most existing point cloud based 3D object detectors focus on the tasks of classification and box regression. However, another bottleneck in this area is achieving an accurate detection confidence for Non-Maximum Suppression (NMS) post-processing. In paper, we add a IoU prediction branch to regular regression branches. The predicted used as NMS. order obtain more prediction, propose IoU-Net with sensitive feature learning alignment operation. To perspective-invariant head, Attentive Corner...
With the increasing seriousness of environmental and energy shortage problems, new technologies such as photovoltaic distributed generations storage devices will be development trend future rural distribution network. In this paper, a bi-layer optimal planning model network is established, where problem on long time scale solved in upper-layer model, operation short lower-layer model. Then, operating converted added into constraints according to Karush–Kuhn–Tucker (KKT) conditions, Big-M...
In this paper, we propose a Monocular 3D Single Stage object Detector (M3DSSD) with feature alignment and asymmetric non-local attention. Current anchor-based monocular detection methods suffer from mismatching. To overcome this, two-step approach. the first step, shape is performed to enable receptive field of map focus on pre-defined anchors high confidence scores. second center used align features at 2D/3D centers. Further, it often difficult learn global information capture long-range...