search

autonomousvision / kitti360Scripts

This repository contains utility scripts for the KITTI-360 dataset.
MIT License
380 stars 61 forks source link

3d bounding boxes format mapping #89

Closed psriram2 closed 1 year ago

psriram2 commented 1 year ago

Hi,

Thank you for your great work. I am quite confused as to how the 3d bounding boxes are mapped to the individual frames. I'm attempting to plot all the 3d bounding boxes given a specific frame (image) from a sequence and I am not sure how to retrieve the 3d bounding boxes corresponding to that frame. Thanks for your help!

yiyiliao commented 1 year ago

Hi, thank you for your interest in the dataset! You can retrieve the 3D bounding boxes of each frame based on the semanticId and instanceId. We provide semanticId and instanceId of each 3D bounding box in the data_3d_bboxes, this allows for retrieving the corresponding objects from the instance labels in the data_2d_semantics. We provide a script for this: https://github.com/autonomousvision/kitti360Scripts/blob/1e5d6371440c6c1b6d2e6d86d9a2869fe5ced45c/kitti360scripts/helpers/annotation.py#L318-L347

You can try it out via the kitti360Viewer.

psriram2 commented 1 year ago

great, thank you! this worked for me - closing this issue