Shape Completion Toolkit

To speed up the usage of our dataset, we provide a development kit providing a PyTorch-based data loader and a small library for computing the metrics that we use in the challenge relative to this dataset. For more details on the dataset you can read our paper link

Dataloader

from torch.utils.data import DataLoader
from dataloader import ShapeCompletionDataset

shape_completion_dataset = ShapeCompletionDataset(data_source='path_to_data/')
dataloader = DataLoader(shape_completion_dataset)

for item in dataloader:
    <do stuff>

Metrics

import open3d as o3d

from chamfer_distance import ChamferDistance
from precision_recall import PrecisionRecall

cd = ChamferDistance()
pr = PrecisionRecall(min_t=0.001,max_t=0.01,num=10)

# list of prediction and corresponding groundtruth
predictions = [...]
groundtruths = [...]

for prediction, groundtruth in zip(predictions,groundtruths):

    cd.update(groundtruth, prediction)  
    pr.update(groundtruth, prediction)  

final_cd_metric = cd.compute()
final_pr_metrics = pr.compute()

cd.reset()
pr.reset()

How to Cite

If you use this repo, please cite as:

@inproceedings{magistri2025icra,
author = {F. Magistri and T. L\"abe and E. Marks and S. Nagulavancha and Y. Pan and C. Smitt and L. Klingbeil and M. Halstead and H. Kuhlmann and C. McCool and J. Behley and C. Stachniss},
title = {{A Dataset and Benchmark for Shape Completion of Fruits for Agricultural Robotics}},
booktitle = icra,
year = 2025,
}