Method

EfficientPS+MaskMatch [EffPS_MM]


Submitted on 9 Oct. 2021 08:10 by
Rohit Mohan (University of Freiburg)

Running time:0.5 s
Environment:1 core @ 2.5 Ghz (C/C++)

Method Description:
Our approach consists of the Effi-
cientPS learning architecture for panoptic
segmentation and anappearance-based instance
association for tracking the pixelsbelonging
to instances. EfficientPS is comprised of a
sharedbackbone, two task-specific heads for
semantic segmentationand instance segmentation,
and a fusion module that generatesthe final
panoptic segmentation output. The backbone
is amodified RegNet model as the encoder,
followed by the 2-way FPN to learn
semantically rich multi-scale features.
Theinstance head consists of a modified
Mask R-CNN architec-ture while the semantic
segmentation head processes featuresat
different scales, allowing coherent feature
refinement. To assign consistent IDs to the
pixels corresponding to the sameinstance
across frames we utilize an appearance-based
matching algorithm.
Parameters:
-
Latex Bibtex:

Detailed Results

From all 29 test sequences, our benchmark computes the STQ segmentation and tracking metric (STQ, AQ, SQ (IoU)). The tables below show all of these metrics.


Benchmark STQ AQ SQ (IoU)
KITTI-STEP 62.93 % 61.49 % 64.41 %

This table as LaTeX




eXTReMe Tracker