| CARVIEW |
Select Language
HTTP/2 200
server: GitHub.com
content-type: text/html; charset=utf-8
last-modified: Mon, 02 Oct 2023 06:39:55 GMT
access-control-allow-origin: *
strict-transport-security: max-age=31556952
etag: W/"651a65bb-106a"
expires: Mon, 29 Dec 2025 05:12:52 GMT
cache-control: max-age=600
content-encoding: gzip
x-proxy-cache: MISS
x-github-request-id: 8544:1387E:83DEC6:94578A:69520B7B
accept-ranges: bytes
age: 0
date: Mon, 29 Dec 2025 05:02:52 GMT
via: 1.1 varnish
x-served-by: cache-bom-vanm7210075-BOM
x-cache: MISS
x-cache-hits: 0
x-timer: S1766984572.912056,VS0,VE220
vary: Accept-Encoding
x-fastly-request-id: 45b19516d3ea68e1841f8c5e043683dd2a1dfd26
content-length: 1686
Equivariant 4D Panoptic Segmentation | Rotation-equivariance brings better performance and efficiency.
Skip to the content.
Equivariant 4D Panoptic Segmentation
Rotation-equivariance brings better performance and efficiency.
This is the project page for the ICCV 2023 paper:
4D Panoptic Segmentation as Invariant and Equivariant Field Prediction
Our model ranks 1st on the SemanticKITTI 4D Panoptic Segmentation public leaderboard!
Two models are presented in our paper: EQ-4D-STOP and EQ-4D-PLS.
Code open-sourced!
EQ-4D-StOP: this link.
EQ-4D-PLS: coming soon.
TL;DR
Main idea
We view the instance clustering strategies through the lens of invariant and equivariant field regression, yielding the two models.
Qualitative result
Take EQ-4D-StOP as an example, the regressed offsets from the equivariant model more accurately point to the instance centers.
Quantitative result
Our equivariant models achieve higher accuracy (LSTQ) with lower memory consumption and running time compared with the non-equivariant baseline.
