论文笔记:Reinforcing LiDAR-Based 3D Object Detection with RGB and 3D Information
2019-10-23 19:18
1616 查看
版权声明:本文为博主原创文章,遵循 CC 4.0 BY-SA 版权协议,转载请附上原文出处链接和本声明。
本文链接:https://blog.csdn.net/weixin_39952791/article/details/102709041
论文笔记:Reinforcing LiDAR-Based 3D Object Detection with RGB and 3D Information
简介
本文提出了一个融合了RGB-D和3D点云信息的3-D目标检测方法。融合了RGB-D的方法在速度有所下降的情况下准确率得到了一定的提高。
方法
这是一个三阶段方法。
第一阶段,用PointRCNN的第一阶段提取点云的3D Proposal
第二阶段,用PointRCNN的第二阶段来refine点云的ROI和预测confidence。
第三阶段,考虑到回归后的3D box信息在第二阶段并没有用在分类上,因此用第二阶段中得到的3D boxes训练一个3D分类器和一个RGB-D分类器(RGB-D分类器用的2D RoIs用3D bounding box投影得到),
因为3D点云提供的信息比较有限,有一些在点云中很容易被分类错误的背景在RGB-D图像中很容易区分,因此最终的分数还加上了第三阶段的分类分数,提高整体的准确率。
相关文章推荐
- 【论文笔记】Region-based Convolutional Networks for Accurate Object Detection and Segmentation
- [论文笔记] Joint 3D Proposal Generation and Object Detection from View Aggregation
- Region-based Convolutional Networks for Accurate Object Detection and Segmentation----R-CNN论文笔记
- 【论文翻译】Road vehicle localization with 2D push-broom LIDAR and 3D priors 论文学习笔记
- 论文笔记 R-FCN: Object Detection via Region-based Fully Convolutional Networks
- 论文笔记|Rich feature hierarchies for accurate object detection and semantic segmentation
- 形变目标识别DPM论文《Object Detection with Discriminatively Trained Part Based Models》阅读笔记
- 论文笔记 R-FCN: Object Detection via Region-based Fully Convolutional Networks
- Rich feature hierarchies for accurate object detection and semantic segmentation论文笔记
- 论文笔记 | Exploit All the Layers: Fast and Accurate CNN Object Detector with SDP and CRC
- 【论文笔记】(CVPR2016 Oral) Training Region-based Object Detectors with Online Hard Example Mining
- 论文笔记:Faster R-CNN:Towards Real-Time Object Detection with Region Proposal Networks
- 论文笔记——Rich feature hierarchies for accurate object detection and semantic segmentation
- R-FCN: Object Detection via Region-based Fully Convolutional Networks 论文笔记
- 论文笔记:Faster R-CNN:Towards Real-Time Object Detection with Region Proposal Networks
- 论文笔记 《Rich feature hierarchies for accurate object detection and semantic segmentation》
- 论文笔记(1)DenseBox: Unifying Landmark Localization with End to End Object Detection
- 论文笔记:Training Region-based Object Detectors with Online Hard Example Mining
- 论文阅读笔记:Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
- 论文笔记| Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks