论文阅读笔记:(2021.06 cvpr) Categorical Depth Distribution Network for Monocular 3D Object Detection

paper:https://openaccess.thecvf.com/content/CVPR2021/papers/Reading_Categorical_Depth_Distribution_Network_for_Monocular_3D_Object_Detection_CVPR_2021_paper.pdf论文阅读笔记:(2021.06 cvpr) Categorical Depth Distribution Network for Monocular 3D Object Detectionhttps://openaccess.thecvf.com/content/CVPR2021/papers/Reading_Categorical_Depth_Distribution_Network_for_Monocular_3D_Object_Detection_CVPR_2021_paper.pdf

code:GitHub - TRAILab/CaDDN: Categorical Depth Distribution Network for Monocular 3D Object Detection (CVPR 2021 Oral)论文阅读笔记:(2021.06 cvpr) Categorical Depth Distribution Network for Monocular 3D Object Detectionhttps://github.com/TRAILab/CaDDN精度对比:KITTI Cars Moderate Benchmark (Monocular 3D Object Detection) | Papers With Code

论文阅读笔记:(2021.06 cvpr) Categorical Depth Distribution Network for Monocular 3D Object Detection

 

主要观点/贡献:

1. pseudo lidar lidar系列的方法, 直接用预测出来的伪点云, 可能会有over confident的问题,尤其对远距离的情形特别明显;

2. 直接把feature投影到3D空间,在预测深度和3D框的方法,存在特征混淆(feature smearing)的问题,检测结果也不好;

3. 因此, 提出了CaDDN: 先预测置信度分布,再把feature投影到3D,再做检测,提高了检测的准确性;

实现:

论文阅读笔记:(2021.06 cvpr) Categorical Depth Distribution Network for Monocular 3D Object Detection

1. 产生voxel feature: a) 用depth distribution把front view的features投影成为frustum features(左下图)  -》 b) 用空间线性插值(trilinear interpolation)把frustum features 差值为voxel features(右下图);

论文阅读笔记:(2021.06 cvpr) Categorical Depth Distribution Network for Monocular 3D Object Detection

2. 按照类似point pillars的方式进行3D目标检测

消融实验:

论文阅读笔记:(2021.06 cvpr) Categorical Depth Distribution Network for Monocular 3D Object Detection

对精度有较大提高的地方:

joint depth supervision with 3d-object detection;

LID;

use full distribution to generate frustum features;

重要参考文献:

1. Center3d: Center-based monocular 3d object detection with joint depth understanding

论文阅读笔记:(2021.06 cvpr) Categorical Depth Distribution Network for Monocular 3D Object Detection

上一篇:如何从MySQL中的字段中仅选择第一个不同的匹配?


下一篇:PowerToys 10.0 Crack