SLIDE 1
Learning to Optimally Segment Point Clouds
Peiyun Hu, David Held, Deva Ramanan Carnegie Mellon University
Paper ID: 2977
Raw LiDAR Scans
Today, most autonomous vehicles perceive the world through LiDAR point clouds.
Learning to Optimally Segment Point Clouds Peiyun Hu, David Held, - - PowerPoint PPT Presentation
Learning to Optimally Segment Point Clouds Peiyun Hu, David Held, - - PowerPoint PPT Presentation
Learning to Optimally Segment Point Clouds Peiyun Hu, David Held, Deva Ramanan Carnegie Mellon University Paper ID: 2977 Raw LiDAR Scans Today, most autonomous vehicles perceive the world through LiDAR point clouds. Map-based Preprocessing
SLIDE 2
SLIDE 3
Map-based Preprocessing
*We focus on a limited field of view in this work.
They often use pre-built maps to first filter out points from the background, then run clustering on points from the foreground to obtain object-level perception.
SLIDE 4
Object-level Perception
However, it is often hard to set the right hyper-parameters for clustering. For example, Euclidean Clustering with a large distance threshold tends to under-segments pedestrians.
*Colors flicker because the algorithm does not track objects across time.
SLIDE 5
Object-level Perception
And Euclidean Clustering with a small distance threshold tends to over-segments vehicles.
*Colors flicker because the algorithm does not track objects across time.
SLIDE 6
No One-Fits-All Solution
= 2.0m
ε
= 1.0m
ε
= 0.5m
ε
= 0.25m
ε
*Colors across parameters do not indicate correspondence.
The best distance threshold often varies from scenario to scenario.
SLIDE 7
A Hierarchical Perspective
= 2.0m
ε
= 1.0m
ε
= 0.5m
ε
= 0.25m
ε
Segmentations with different thresholds form a hierarchy, where nodes represent segments.
SLIDE 8
Learning Objectness Models
PointNet++ (Qi et al., NeurIPS’17) We learn a model to predict an objectness score for each segment in the hierarchy.
Bad Good
0.9 0.1
SLIDE 9
Objectness
How well a segment overlaps with ground truths.
SLIDE 10
Searching for Optimality
Given a hierarchy of segments with scores, we search for the optimal segmentation.
Bad Good
SLIDE 11
Bad Good
defines
Optimal Worst-case Segmentation
We propose an efficient algorithm that produces optimal segmentation under this definition.
the worst segment score segmentation score
SLIDE 12
Bad Good
defines
Average-case Segmentation
We also propose an efficient algorithm guided by average-case score.
average local segment score global segmentation score
SLIDE 13
Quantitative Evaluation
Protocol: compute the percentage of objects that are under-segmented and over-segmented.
U = 1 L
L
∑
l=1
1[ |Ci* ∩ Cgt
l |
|Ci*| < τU]
2 under-segmented pedestrians 1 over-segmented car
O = 1 L
L
∑
l=1
1[ |Ci* ∩ Cgt
l |
|Cgt
l |
< τO]
Assumption: output is a valid partition.
Held et al., RSS’15
SLIDE 14
Segmentation Errors
(1) As distance threshold increases, more under-segmentation and less over-segmentation. (2) Our adaptive algorithm significantly outperforms each single-parameter baseline. (3) We also plot the lower-bound errors for the search space, showing room for improvement. 0.25 0.5 0.75 1 Under Over Total
13% 5% 8% 17% 8% 9% 19% 6% 13% 28% 5% 23% 35% 25% 9% 68% 65% 3% 92% 91% 1%
CC(0.25m) CC(0.5m) CC(1m) CC(2m) Ours(min) Ours(avg) CC(*)
SLIDE 15
No One-Fits-All Solution
= 2.0m
ε
= 1.0m
ε
= 0.5m
ε
= 0.25m
ε
*Colors across parameters do not indicate correspondence.
The best distance threshold often varies from scenario to scenario.
SLIDE 16
Algorithmic Output
Our algorithm can adaptively choose the best distance threshold for each scenario.
SLIDE 17