Skip to content
JournalsWorldThe Global Research Discovery Platform
Featured Dataset

Re-assembling the past: The RePAIR dataset and benchmark for real world 2D and 3D puzzle solving

Accepted by NeurIPS 2024 Datasets and Benchmarks Track We introduce the RePair puzzle-solving dataset, a large-scale real world dataset of fractured frescoes from the archaelogical campus of Pompeii. Our dataset consists of over 1000 fractured frescoes. The RePAIR stands as a real

👤
CreatorTsesmelis, Theodore
📅
Published2025-07-03
🔗
DOI10.5281/zenodo.15800029
📊
Downloads5,981
File Size44.6 GB
Data TypeDataset
Published2025
Total Views2,540
Total Downloads5,981

Accepted by NeurIPS 2024 Datasets and Benchmarks Track

We introduce the RePair puzzle-solving dataset, a large-scale real world dataset of fractured frescoes from the archaelogical campus of Pompeii. Our dataset consists of over 1000 fractured frescoes. The RePAIR stands as a realistic computational challenge for methods for 2D and 3D puzzle solving, and serves as a benchmark that enables the study of fractured object reassembly and presents new challenges for geometric shape understanding. Please visit our website for more dataset information, access to source code scripts and for an interactive gallery viewing of the dataset samples.

Access the entire dataset

We provide a compressed version of our dataset in two seperate files. One for the 2D version and one for the 3D version.

Our full dataset contains over one thousand individual fractured fragments divided into groups with its corresponding folder and all compressed into their individual sub-set format regarding whether they are 2D or 3D. Regarding the 2D dataset, each fragment is saved as a .PNG image and each group has the corresponding ground truth transformation to solve the puzzle as a .TXT file. Considering the 3D dataset, each fragment is saved as a mesh using the widely .OBJ format with the corresponding material (.MTL) and texture (.PNG) file. The meshes are already in the assembled position and orientation, so that no additional information is needed.  All additional metadata information are given as .JSON files.

 

Important Note

Please be advised that downloading and reusing this dataset is permitted only upon acceptance of the following license terms.

The Istituto Italiano di Tecnologia (IIT) declares, and the user (“User”) acknowledges, that the “RePAIR puzzle-solving dataset” contains 3D scans, texture maps, rendered images and meta-data of fresco fragments acquired at the Archaeological Site of Pompeii. IIT is authorised to publish the RePAIR puzzle-solving dataset herein only for scientific and cultural purposes and in connection with an academic publication referenced as Tsemelis et al., “Re-assembling the past: The RePAIR dataset and benchmark for real world 2D and 3D puzzle solving”, NeurIPS 2024. Use of the  RePAIR puzzle-solving dataset by User is limited to downloading, viewing such images; comparing these with data or content in other datasets. User is not authorised to use, in particular explicitly excluding any commercial use nor in conjunction with the promotion of a commercial enterprise and/or its product(s) or service(s), reproduce, copy, distribute the  RePAIR puzzle-solving dataset. User will not use the  RePAIR puzzle-solving data

📤 Share this page

Found this useful? Share it with your network.

✓ Link copied! Paste it on ResearchGate / Academia.edu
📦
Re-assembling the past: The RePAIR dataset and benchmark… (Full Dataset)44.6 GB
⬇
📄
ReadmeVia DOI record
↗

Files are hosted on the source repository. Click download to access the full dataset.

Tsesmelis, Theodore (2025). Re-assembling the past: The RePAIR dataset and benchmark for real world 2D and 3D puzzle solving. https://doi.org/10.5281/zenodo.15800029