search

happinesslz / EPNet

EPNet: Enhancing Point Features with Image Semantics for 3D Object Detection(ECCV 2020)
MIT License
234 stars 36 forks source link

Model input and output #21

Closed niujiaxing closed 3 years ago

niujiaxing commented 3 years ago

https://github.com/happinesslz/EPNet/blob/0123c341243846aa3b412addcb9e2c07fd305237/tools/eval_rcnn.py#L549 Hello author, I want to see the format of the model input and output print, if you have time,could you explain what these variables represent, thank you very much 1.txt

happinesslz commented 3 years ago

@niujiaxing

input: 'pts_input': the input point cloud for Geometric Stream in Two stream RPN. 'pts_origin_xy': the corresponding coordinates on the image from 3D point cloud to 2D image. 'img': the input image for Image Stream in Two stream RPN.

output: 'rpn_cls' and 'rpn_reg' : the classification and regression in Two stream RPN, respectively. 'backbone_xyz' and 'backbone_features': the coordinates and their corresponding features of point cloud, respectively. 'rois': the output 3D boxes in RPN stage, which will be used in RCNN stage. 'seg_result': the segmentation in Two stream RPN, which can be obtained from 'rpn_cls'. 'rcnn_cls' and 'rcnn_reg': the classification and regression in RCNN stage, respectively.