Here are some frequently used datasets in my papers
Here are some frequently used datasets in my papers
Data: gray scale images of handwritten digits (MNIST) or characters (EMNIST). Each image is of size 28-by-28.
Scattering transform extracts invariant and stable image representations. Subspace clustering on such representations produces state-of-the-art performance.
Usage from Matlab: to extract features, follow the instructions in the package [here]
Usage from Python: this has been made considerably easier by the Kymatio package. See an example of usage [here]
Related papers that use these datasets:
Oracle Based Active Set Algorithm for Scalable Elastic Net Subspace Clustering, CVPR'16
A Scalable Exemplar-based Subspace Clustering Algorithm for Class-Imbalanced Data, ECCV'18
Is an Affine Constraint Needed for Affine Subspace Clustering? ICCV'19
Data: gray scale face images of 38 subjects each taken under around 64 illumination conditions. Each image is of size 192-by-168
We put together a .mat file for the entire dataset [download]. The images are down-sampled to 48-by-42 via bicubic interpolation.
Usage from Matlab:
% EYALEB_DATA: 2016-by-2414 matrix where each column is a flattened face image
% EYALEB_LABEL: 1-by-2414 vector of corresponding labels, taken values in 1, ..., 38
% EYALEB_NAME: 1-by-2414 cell of file names
load ExtendedYaleB.mat EYALEB_DATA EYALEB_LABEL EYALEB_NAME
% Images can be visualized as imshow(uint8(reshape(EYALEB_DATA(:, 1), [48 42])))
Usage from Python: use the code [here] together with torch.utils.data.DataLoader
Related papers that use this dataset:
Oracle Based Active Set Algorithm for Scalable Elastic Net Subspace Clustering, CVPR'16
Scalable Sparse Subspace Clustering by Orthogonal Matching Pursuit, CVPR'16
Self-Supervised Convolutional Subspace Clustering Network, CVPR'19