Novel Artistic Scene-Centric Datasets for Effective Transfer Learning in Fragrant Spaces

TL;DR

This paper introduces new datasets and a transfer-learning method to improve classification of fragrant spaces and artistic scenes, leveraging weakly labeled data from cultural heritage sources.

Contribution

The work presents novel datasets and demonstrates that transfer learning with weak supervision significantly enhances classification accuracy in artistic scene recognition.

Findings

Transfer learning improves classification of fragrant spaces.

Weakly labeled data from cultural heritage sources is effective.

The ArtPlaces dataset supports further research.

Abstract

Olfaction, often overlooked in cultural heritage studies, holds profound significance in shaping human experiences and identities. Examining historical depictions of olfactory scenes can offer valuable insights into the role of smells in history. We show that a transfer-learning approach using weakly labeled training data can remarkably improve the classification of fragrant spaces and, more generally, artistic scene depictions. We fine-tune Places365-pre-trained models by querying two cultural heritage data sources and using the search terms as supervision signal. The models are evaluated on two manually corrected test splits. This work lays a foundation for further exploration of fragrant spaces recognition and artistic scene classification. All images and labels are released as the ArtPlaces dataset at https://zenodo.org/doi/10.5281/zenodo.11584328.