Skip to content

Dataset · chrischoy

3DMatch point clouds for FCGF training

chrischoy/FCGF-3DMatch

Preprocessed 3DMatch 3D scene fragments and their overlap lists for geometric feature descriptor training.

What's inside

Inside are preprocessed point clouds in .npz format (1,000 to 10,000 records) and .txt text files with lists of overlapping fragment pairs and their intersection degree. The data was prepared using the official script from the FCGF repository (ICCV 2019 paper) for 3D space matching.

What you can do with it

  • Training 3D feature extraction models (e.g., FCGF) for point cloud matching
  • Testing 3D registration and scene fragment stitching algorithms
Tasks
Other
Size
1K–10K rows
Downloads this month
20.6K
Likes
0

How to load it

from datasets import load_dataset

ds = load_dataset("chrischoy/FCGF-3DMatch")