EEG signals for Object Segmentation

Authors: Eva Mohedano*, Graham Healy*, Kevin McGuinness*, Xavier Giró-i-Nieto, Noel E. O'Connor*, Alan F. Smeaton*.

(*) Insight Center of Data Analytics, Dublin City University (Ireland)

 This dataset contains the data used in 'Classification of EEG Signals for Object Segmentation' (submitted). It contains the 22 images presented to the users and the EEG data, recordered during the experiments. The time of the visual events is provided with EEG recorders. Also, the extracted EEG epochs associated to the visual events, with their corresponded labels for its classification.

IMAGE DATASET (61 MB) [Dataset.zip]:

It contains 22 images for object segmentation.The dataset includes different configurations regarding the color, shape, and texture of the objects, as well as their relative similarity with the foreground. The collection consists of 20 new images captured for the purpose of this work and images 38082 and 123074 from the Berkeley Segmentation Dataset and Benchmark (BSDB) [1]. Each of the images has an associated ground truth in the form of a binary mask. In the case of the two BSDS images, the ground truth masks were obtained from a previous work where 100 binary masks from objects where generated from a subset of 96 images published by the Dublin City University [2]. Each image has been cropped into 192 windows for its presentation. Information about the number of object pixels that appears in each window is provided.

EEG DATASET (1.6GB) [Data_EEG.tar]:

It contains the EEG data recorded from a non-invasive 31 channel BCI with a sample rate of 1kHz, with the electrodes were located according to the 10-20 system distribution for the recordings. 

There are provided the recorders from 5 users, for each one it is provided the .vhdr, .eeg and .vmrk files with the data.

EPOCH DATASET (3.7GB) [Users_epochs.tar]:

It contains the labeled epochs for the visual stimulus presented. The epochs are segments from the original EEG dataset related to each window presentation (1 and 2 seconds before and after the window presentation), after applying a downsapling to 250Hz and a band-pass filter from 0.1Hz to 70Hz.

[1] Martin, David, et al. "A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics." Computer Vision, 2001. ICCV 2001. Proceedings. Eighth IEEE International Conference on. Vol. 2. IEEE, 2001. 

[2] McGuinness, Kevin, and Noel E. O’Connor. "A comparative evaluation of interactive segmentation algorithms." Pattern Recognition 43.2 (2010): 434-444.