metadata
license: cc-by-4.0
task_categories:
- image-segmentation
tags:
- boundary-detection
- computer-vision
- stereo-vision
size_categories:
- n<1K
Multi-Cue Boundary Detection Dataset
Official dataset release by Serre Lab, Brown University.
A dataset of 100 stereo video sequences for studying the interaction of multiple early visual cues (luminance, color, stereo, motion) during boundary detection in natural scenes.
Dataset Specifications
- Scenes: 100 natural scenes
- Format: Binocular video sequences (left and right views)
- Sequence length: 10 frames per scene
- Frame rate: 30 fps
- Resolution: 1280 x 720 pixels
Files
- multicue-dataset.tar.bz2 — Full dataset (~3.4 GB)
Citation
@article{mely2016systematic,
title={A systematic comparison between visual cues for boundary detection},
author={M{\'e}ly, David A and Kim, Junkyung and McGill, Mason and Guo, Yihan and Serre, Thomas},
journal={Vision Research},
volume={120},
pages={93--107},
year={2016}
}