branch | description | owner | Update |
---|---|---|---|
master | fork from Jerrypiglet/kitti_instance_RGBD_utils | youyi | |
test_heatmap | run testing | youyi | |
dump_data | dump TUM dataset using kitti scripts | youyi |
A data preparation script for instance-wise temporal RGB-D/3D data on KITTI.
Each sample consists of one car sequence with:
- N RGB frames cropped out;
- N reprojected sparse depth from KITTI raw data;
- Style and pose initialization from MV3D dataset for the first frame in the sequence;
- 2D/3D bounding boxes;
- Silhouette(s) for some frames (not for all samples because only 200 frames of KITTI are semantically labelled).
KITTI raw, depth, semantic, detection dataset; MV3D dataset.
Mayavi(http://docs.enthought.com/mayavi/mayavi/mlab.html) is required to run the code.
Quick installation in Python2 with Conda:
https://stackoverflow.com/questions/41960672/how-to-install-mayavi-trait-backends
Check the kitti_lidar_reproj_ipynb.ipynb
file for a demo.
python kitti_lidar.py --fdir [input dir of kitti drive] --outdir [output dir]
Example:
conda activate mayavi
python kitti_lidar.py --fdir /home/rzhu/Documents/kitti_dataset/raw/2011_09_26/2011_09_26_drive_0005_sync/ --outdir ./