SynGallery: A Synthetic Gallery of Real Paintings for Instance-Level Artwork Recognition

Patryk Bartkowiak1,5, Jakub Markil1, Bartosz Kotrys2, Dominik L. Michels3,5, Sören Pirk4,5, Wojtek Palubicki1,5
1Adam Mickiewicz University, 2ArtiCollect, 3KAUST, 4Kiel University, 5GreenMatterAI
VISART @ ECCV 2026
arXiv Code Dataset HF Collection Poster
Cavalier and Shepherd rendered at a 150 degree viewing angle
150°
Cavalier and Shepherd rendered at a 120 degree viewing angle
120°
Cavalier and Shepherd rendered at a 90 degree viewing angle
90°
Cavalier and Shepherd rendered at a 60 degree viewing angle
60°
Cavalier and Shepherd rendered at a 30 degree viewing angle
30°

SynGallery places each painting in a procedurally generated 3D gallery and renders it from many viewpoints while preserving its identity. Above: Cavalier and Shepherd by Francesco Casanova, seen from five viewpoints.

4,898real paintings from the Met
24,490rendered gallery views
73.47GAP⁻ training on renders, against 67.18 on studio photos at the same budget
38.48GAP on the full Met benchmark with SynGallery added, up from 35.97

Abstract

Instance-level artwork recognition requires matching a handheld visitor photograph to a specific work in a large museum collection. This is challenging because painting datasets typically provide clean catalog images for training, while test queries are captured under oblique viewpoints, gallery lighting, reflections, frames, and other scene-level variations. We present SynGallery, a synthetic gallery dataset for artwork retrieval that addresses this gap without collecting additional real photographs. Starting from catalog images of real paintings, we place each artwork into a procedurally generated 3D gallery scene and render it from multiple viewpoints under varied geometric and appearance conditions, while preserving the exact identity of the original work.

The resulting dataset contains 24,490 rendered views of 4,898 paintings from the Met benchmark. We show that these synthetic views provide a stronger training signal than the corresponding studio photographs. At the same number of training data points, training only on SynGallery improves art painting recognition from 67.18 to 73.47 GAP−. When added to the full Met training set, SynGallery improves the published benchmark protocol from 35.97 to 38.48 GAP. Ablation experiments show that the gain comes from scene-level view variation rather than photographic realism: reducing the five rendered viewpoints to a single frontal view removes most of the improvement, while simulating capture artifacts such as blur, sensor noise, and image compression consistently reduces performance.

Method Overview

Overview of the SynGallery pipeline: a catalog image of a real painting is placed into a procedurally generated 3D gallery scene and rendered from multiple viewpoints under varied conditions, preserving the painting's identity.

From a single catalog image, SynGallery places each painting in a procedurally generated 3D gallery and renders it from multiple viewpoints under varied geometric and appearance conditions, while preserving the exact identity of the original work.

Renders beat studio photos

Replace the real catalog photos with SynGallery renders at the same number of images, then add more renders. Every step towards synthetic data helps.

Retrieval quality by training data
Hover a point for the numbers; pick the metric.

Viewpoints matter most

Varying one generation parameter at a time shows where the gain comes from: geometric variation, not photographic realism.

Dataset ablation, closed painting world
Every model is trained on renders only.

Sample Renders

Acknowledgments

ArtiCollect EOSC-ARENA Poznańskie Centrum Superkomputerowo-Sieciowe (PCSS)

This research was conducted in collaboration with ArtiCollect, who funded the work, contributed to data collection, and provided domain expertise in art.

This work was supported by the European Union through the EOSC-ARENA project (AI Research Enhancement through Networked Agents), Grant Agreement No. 101292597.

The authors thank Poznańskie Centrum Superkomputerowo-Sieciowe (PCSS) for providing the computational resources used in the experiments.

BibTeX

@misc{bartkowiak2026syngallery,
      title={SynGallery: A Synthetic Gallery of Real Paintings for Instance-Level Artwork Recognition},
      author={Patryk Bartkowiak and Jakub Markil and Bartosz Kotrys and Dominik Michels and Sören Pirk and Wojtek Palubicki},
      year={2026},
      eprint={2607.18907},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2607.18907},
      note={VISART @ ECCV 2026},
}