Hybrid Relative Attributes Based on Sparse Coding for Zero-Shot Image Classification

Algorithm 2

Zero-shot image classification based on SC-HRA.

Input: the attribute ranking scores of the training and testing samples

Output: the label of the testing sample

Building of the training models;

Calculate mean value and covariance matrix based on of training sample,

obtain

Building of the testing models;

If is satisfied, then the mean value of model is

, and the covariance matrix is .

If is satisfied, then the mean value of model is , and

the covariance matrix is .

If is satisfied, then the mean value of model is , and

the covariance matrix is .

Testing;

By calculating the attribute ranking score of the testing sample, the class label is

determined as