Hand-drawn symbols in VR

Hand-drawn symbols in VR database (Matlab format)

Data Set Name: HDSVR

Hand-drawn symbols (64x64) in virtual reality.

• • Data type: multivariate

  • Task: classification

  • Attribute type: real numbers

  • Area: Document analysis and symbol recognition

  • Number of instances: 3059

  • Number of features: 4096 (64x64)

The dataset contains 3059 images of size 64x64 that represent hand drawn symbols (30 classes) that were acquired within an immersive virtual reality environment representing a white board, using an HTC Vive and its controllers. This dataset represents an example of pattern recognition for virtual reality.

The format of the database is in Matlab:

• • The images: img_data(64x64x3059)

  • The groundtruth gt_data(1x3059).

The image are already normalized [0-1]. In addition, we provide a database of the original symbols that were used as a model.

Check the number of images per class:

load('data_symbols.mat');

l=unique(gt_data);

nclasses=size(l,1);

for i=1:nclasses

nimg(i)=size(find(gt_data==l(i)),1);

end

Attribute information:

• • Images of size 64*64 representing symbols.

  • Minimum number of examples for a class: 96

  • Maximum number of examples for a class: 124

Citation Requests / Acknowledgements:

Please cite this paper if you use the database:

• • H. Cecotti, C. Boumedine, M. Callaghan, Hand-drawn symbol recognition in immersive virtual reality using deep extreme learning machines, Int. Conf. on Recent Trends in Image Processing and Pattern Recognition (LNCS Communications in Computer and Information Science vol. 709 ), pp. 80-92, 2017.