MSTA3D - Multi-scale Twin-attention for 3D Instance Segmentation

July 21, 2024

2024   ·   images   instance-segmentation   transformer   visualization     ·   paper  

Visualize results on validation set of ScanNetv2 dataset.

Top row: from left to right are Input Point Cloud, High-scale Superpoint, Low-scale Superpoint and Ground-truth respectively
Bottom row: from left to right are ISBNet, SPFormer, MAFT and MSTA3D respectively

Other examples

1 / 4
ISBNet
2 / 4
SPFormer
3 / 4
MAFT
4 / 4
MSTA3D


Enjoy Reading This Post?

Here are some more posts you might like to read next:

  • Mobile Robot
  • 3DoF Arm Robot