Algorithm data for the 2017 aesthetic value project (NESP TWQ 3.2.3, Griffith Institute fo

2,500 manually selected images of the Great Barrier Reef, each scored for aesthetic value by at least 10 participants in an online survey launched in October 2017. The dataset includes the original and resized images, survey results in an Excel file, and a full deep learning framework setup for training and testing aesthetic prediction models. It was prepared by the Griffith Institute for Tourism Research as part of the NESP TWQ 3.2.3 project.

Use Cases

• Training deep learning models for automatic image aesthetic scoring based on the provided 2,500 labeled images.
• Benchmarking aesthetic prediction algorithms against human survey scores contained in the Qualtrics.xls file.
• Replicating or extending the original research methodology using the included Caffe framework code and configuration files.

Strengths

• Includes 2,500 manually curated and scored images, providing a labeled dataset for supervised learning.
• Survey methodology is documented, with 705 qualified questionnaires used to derive average aesthetic scores.
• Provides a complete research artifact, including raw images, processed data in LMDB format, model definitions, and training code.

Limitations

• Description metadata is limited; actual data quality and model performance require manual inspection after download.
• Row count and specific column-level documentation for auxiliary files are absent.
• Last updated 2026-06-27 17:07:42.098580; freshness should be verified.

Provenance

Source

Griffith Institute for Tourism Research, Australian Ocean Data Network

Collection Method

Images downloaded via Flickr API using keyword 'Great Barrier Reef' in August-September 2017, followed by manual selection and an online survey on Qualtrics.

Time Range

2017

Freshness

Last updated date is 2026-06-27, but the data collection occurred in 2017.

Geography

Great Barrier Reef