EVA

Dataset Overview

Dataset Components

• File Types: Includes 5 CSV files and 7 archive files.
• Images: Resized images from the AVA dataset stored in 7 archives named EVA_together.zip.00X, total size approximately 600‑700 MB.
• CSV Files:
  • image_content_category.csv: Contains image IDs and content categories; each image belongs to a single category.
  • users.csv: Records user information, including ID, age, region, photography skill level, gender, and visual acuity.
  • votes.csv: Records user votes and difficulty assessments for images, as well as visual, composition, quality, and semantic attributes.
  • votes_filtered.csv: Contains filtered user votes and attribute assessments.
  • region_index.csv: Provides a list of regions and their codes.

Dataset Usage

• Citation: When using this dataset, cite the paper "EVA: An Explainable Visual Aesthetics Dataset".
• Image Extraction: After merging the 7 archive files, extract the images.
• Image Classification: Can use provided classifications or define custom ones.

User Information

• User CSV: Includes user ID, age, region, photography skill level, gender, and visual acuity.
• Trusted Users: Lists 13 trusted user IDs.

Voting Information

• Votes CSV: Records image ID, user ID, rating, difficulty, and image attributes.
• Filtered Votes: Provides filtered ratings and attribute data.

Region Information

• Region CSV: Supplies region codes and names.