README.md

metadata

license: cc-by-4.0
task_categories:
  - other
tags:
  - eye-tracking
  - saliency
  - visual-attention
  - gaze
  - fixations
  - polygonal-aperture
pretty_name: 'PolyCAT: Polygon-Aperture Eye-Tracking Dataset'
size_categories:
  - 100K<n<1M

PolyCAT: Polygon-Aperture Eye-Tracking Dataset

PolyCAT is a public eye-tracking dataset for studying visual attention on natural images viewed through irregular polygon apertures. The dataset is designed to support saliency prediction, gaze modeling, and research on how geometric viewing constraints affect visual exploration strategies.

Recording Details

• Eye tracker: EyeLink 1000+ (SR Research), head-mounted, binocular, 500 Hz per eye
• Display: 27" 4K monitor (3840 x 2160 pixels) at 144 Hz
• Viewing distance: 70 cm
• Participants: 30 included (ages 24-29, 14 female / 16 male)
• Stimuli: 600 images from CAT2000 (6 categories) viewed through 27 polygon aperture shapes
• Trials per participant: 702 (351 per part x 2 parts)
• Stimulus duration: 4.0 seconds per trial
• Secondary task: Old/new memory probe after each block

Folder Overview

PolyCAT/
├── data/
│   ├── metadata/          # Consolidated tables: participants, sessions, trials, quality
│   ├── gaze/              # Sample-level gaze streams (TSV, one per session)
│   ├── fixations/         # Fixation events (CSV, global + per-participant)
│   ├── saliency_maps/     # Empirical fixation density maps per polygon-image
│   ├── scanpaths/         # Per-trial fixation sequences
│   └── stimuli/           # CAT2000 images and polygon definitions (JSON)
├── code/
│   └── examples/          # Quickstart scripts and notebooks
├── manifests/             # Stimulus manifest, polygon geometry
└── docs/                  # Data dictionary, file formats, acquisition protocol

Getting Started

import pandas as pd

# Load metadata
participants = pd.read_csv("data/metadata/participants.csv")
trials = pd.read_csv("data/metadata/trials.csv")
fixations = pd.read_csv("data/fixations/fixations_all.csv")

# Filter to one participant
p01_fix = fixations[fixations["participant_id"] == "P01"]
print(f"P01 has {len(p01_fix)} fixations across {p01_fix['trial_uid'].nunique()} trials")

# Fixation heatmap for a specific polygon-image combination
import numpy as np
polygon = "polygon_25"
subset = fixations[(fixations["polygon_id"] == polygon) & (fixations["eye"] == "R")]
heatmap = np.zeros((2160, 3840))
for _, row in subset.iterrows():
    x, y = int(row["x_px"]), int(row["y_px"])
    if 0 <= x < 3840 and 0 <= y < 2160:
        heatmap[y, x] += 1

Polygon Apertures

The 27 polygon shapes are organized into three groups:

• Reference (3): regular geometric shapes (rectangle, symmetric, asymmetric)
• Convexity-varied (3): shapes with varying convexity (convex, concave, intermediate)
• Irregular (21): asymmetric polygons with diverse geometric properties

Citation

If you use this dataset, please cite:

@inproceedings{polycat2026,
  title     = {How Do We Look at Images Through Polygonal Apertures: The PolyCAT Dataset},
  author    = {Aklilu, Brahan and Ben-Shahar, Ohad},
  booktitle = {Proceedings of the 2026 ACM Symposium on Eye Tracking Research \& Applications (ETRA '26)},
  year      = {2026},
  doi       = {TBD}
}

License

This dataset is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).