The differences between one_box_est and two_box_est

jacky121298 / 3DAL_PyTorch

This is the pytorch implementation of 3DAL proposed by Qi et. al, "Offboard 3D Object Detection from Point Cloud Sequences", CVPR 2021.

12 stars 2 forks source link

The differences between one_box_est and two_box_est #1

Open Pan-panpan opened 5 months ago

Pan-panpan commented 5 months ago

Hello! Thank you for the code you provided. How should I set the model_type? What are the differences between one_box_est and two_box_est? Are they related to the settings of sweeps?

jacky121298 commented 5 months ago

Hello! Thank you for the code you provided. How should I set the model_type? What are the differences between one_box_est and two_box_est? Are they related to the settings of sweeps?

Hi! The model_type parameter refers to the type of box regression network in the static object auto-labeling model. "one_box_est" denotes the use of a single PointNet-based network to regress the object's bounding box, while "two_box_est" indicates iterative regression using two PointNet-based networks.

Pan-panpan commented 5 months ago

Thank you for your response! I've encountered another issue. I've been using the pre-trained model CenterPoint.pth that you provided, but the performance is quite poor. The bounding box positions are very inaccurate.