search

happinesslz / EPNet

EPNet: Enhancing Point Features with Image Semantics for 3D Object Detection(ECCV 2020)
MIT License
230 stars 37 forks source link

Two questions about the code #22

Closed 304886938 closed 3 years ago

304886938 commented 3 years ago

Hi! I'm learning EPNet recently, your job is great and it is very helpful to me. I have two questions about the code.

  1. In kitti_rcnn_dataset.py, I only find the point cloud has data augmentation but image has not data augmentation, can they align?
  2. In kitti_rcnn_dataset.py, if cfg.LI_FUSION.ENABLED is true, I think point cloud and img are both loaded by the method of self.get_rpn_with_li_fusion, and we don't need to use the method of self.get_rpn_sample, am I right? Look forward to your answers! Thank you very much!
XiwuChen commented 3 years ago
  1. The alignment is utilized before data augmentation and the original pixel coordinate of each point is kept. So they are still aligned.
  2. Yes, you are right. We keep self.get_rpn_sample for ablation study.
304886938 commented 3 years ago

Thank you very, very much for your answer!!!