This article is an original article of the blogger, and may not be reproduced without the permission of the blogger.
This article is a series of articles in the column "python 3D point cloud from basics to deep learning", and the address is "https://blog.csdn.net/suiyingy/article/details/124017716".
For the CenterPoint data and source code configuration and debugging process, please refer to the previous blog post: https://blog.csdn.net/suiyingy/article/details/128002709 . This article mainly introduces the CenterPoint network structure and its running state in detail.
1 Overall process of CenterPoint model
The overall structure of the CenterPoint model is shown in the figure below, from the initial one-stage model to a two-stage model. The second stage is responsible for fine-tuning and correcting the detection results of the first stage, which is basically consistent with the two-stage target detection idea based on candidate boxes. Here we focus on the first stage of CenterPoint, and the single-stage CenterPoint can directly complete the detection of three-dimensional objects.
