---
license: mit
tags:
- print-duration
- 3D-printing
- G-code
- supervised-learning
- regression
- time-series
size_categories:
- 1M<n<10M
language:
- en
pretty_name: 3DTime
---

# 3DTime dataset: (sample of) A Large Dataset of Multivariate Time-Series for 3D-printing Duration

This dataset is a small sample of the 3DTime dataset, for which the paper is currently under review for NeurIPS Datasets and Benchmarks 2026.

This smaller version contains:

- 82 3D models (~0.08% of the full dataset), their corresponding sliced G-code, compressed annotated G-code, and binary vectorized files
- A total of 5,855,369 G-code instructions (~0.07% of the full dataset)

## Croissant metadata file and automatic loading

Due to the peculiar file format, it is not possible to automatically load the actual G-code instruction data using the automatic croissant loading techniques. Instead, use the code provided in the repository linked below.

It is however still possible to load the G-code metadata using those tools:

```python
from datasets import load_dataset
dataset = load_dataset("3DTimeDataset/3DTime")
```

Which will load the dataset G-code level data (such as file names, file sizes, slice time, print time, etc).

IMPORTANT NOTE FOR REVIEWERS:

For now, this metadata loading regards only this smaller version of the dataset (aka the loaded CSV contains 82 rows instead of 99,005). The full dataset, and its corresponding metadata, is still available via the dataset private link in the submission.

We guarantee to make the full dataset public, and to change this metadata automatic loading to the full dataset, upon acceptance.

## Other links

Full dataset access link (NOTE FOR REVIEWERS: this DOI link will be made accessible upon acceptance, in the meantime, the OpenReview submission contains an equivalent private link):

> Masked for review

Code repository (highly recommended in order to use the dataset):

> https://github.com/3DTimeDataset/3DTime_pytorch_dataloader