Submitted by Duc Tran.

Submission data

Full nameTwo-scale Superpoint with Twin-attention for 3D Instance Segmentation
Publication titleMSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation
Publication authorsDuc Tran Dang Trung, Byeongkeun Kang, Yeejin Lee
Publication venueACM Multimedia 2024
Input Data TypesUses Color,Uses Geometry        Uses 3D
Programming language(s)Python
HardwareNVIDIA A100
Websitehttps://arxiv.org/abs/2411.01781
Source code or download URLhttps://github.com/vcl-seoultech/MSTA3D
Submission creation date27 Sep, 2023
Last edited5 Nov, 2024
Last uploaded23 Mar, 2024

3D semantic instance results



Infoavg ap 25%bathtubbedbookshelfcabinetchaircountercurtaindeskdoorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwindow
0.8791.0000.9940.9210.8070.9390.7710.8870.9230.8620.7220.7680.7561.0000.9100.9040.8360.9990.824