Tracking and Visualization for Point Cloud 3D Object Detection

>>> By Deecamp Focus-Det Group <<<

1 Dependency

pip install -r requirements

2 Visualization with tracking

• usage

python visualization_with_tracking.py -dets_dir your_dets_dir -lidar_dir your_lidar_dir -eval_dir your_eval_dir

or simple

python visualization_with_tracking.py -d your_dets_dir -l your_lidar_dir -e your_eval_dir

• description

After obtaining 3-d object detection result based on point cloud, we track the objects by AB3DMOT model frame by frame. To visualize the result, you need to feed the detections directory (dets_dir) which includes some text files (.txt), LiDAR directory (lidar_dir) which includes some point cloud data files (.bin), and tracking result directory (eval_dir) .

3 Visualization after tracking

• usage

python visualization_with_tracking.py -dets_dir your_dets_dir -lidar_dir your_lidar_dir

or simple

python visualization_with_tracking.py -d your_dets_dir -l your_lidar_dir

• description

We track the objects by AB3DMOT model frame by frame after obtaining 3-d object detection result. Then tracking result will be obtained. To visualize the result, you need to feed the detections directory (dets_dir) which includes some text files (.txt), and LiDAR directory (lidar_dir) which includes some point cloud data files (.bin).

4 Format of 3-d object detection result format

0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
type	-	-	-	-	-	-	-	w	l	h	x	y	z	r_y	conf

5 Format of tracking result

0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17
frame	id	type	truncated	occluded	alpha	2d	2d	2d	2d	h	w	l	x	y	z	r_y	conf

6 Example visualization

• shortcut

• shortcut-focus

Reference

• https://github.com/xinshuoweng/AB3DMOT
• http://www.open3d.org/docs/release/