second.pytorch icon indicating copy to clipboard operation
second.pytorch copied to clipboard

How to retrain on my own 16velodyne data

Open khanln opened this issue 6 years ago • 3 comments

Hi, you did a very great job, but when I use your pre-trained model to predict my own 16-vlp data, including retrained kitti data, the result is not good. My velodyne installation height is 1.6m, and did not modify the parameters in the second configuration file. So I hope you can give me some suggestions:

  1. Do I need to modify any configuration to apply to 16-vlp data?

  2. If I want to train my 16-vlp data, how to rewrite a class to modify the data loader, there is no general idea and method(I only have velodyne annotation data).

very appreciate:)

khanln avatar Aug 05 '19 01:08 khanln

@khanln Do you find answer for your questions?

MaAmini avatar Oct 31 '19 18:10 MaAmini

i want to train mu 16-vlp data too ,can you tell me you email

wanghengmm avatar Jan 08 '20 08:01 wanghengmm

@MaAmini @khanln Hi, do you guys find some way to solve this problem?

ryontang avatar Jan 14 '21 13:01 ryontang