Abstract In this paper, we propose a Coupled Discriminative Dictionary Learning framework to tackle the zero-shot image classification problem. Instead of the original attribute vectors and sample feature vectors, we use their corresponding sparse coefficients attained from sparse coding to do the classification. The purpose of our framework is that, when an unseen-class sample shows during test time, we first attain its corresponding sparse coefficient through learned feature dictionary. Then we use a mapping method to map it to the attribute sparse coefficients category histogram domain where we can accomplish the classification. We evaluate our method performance on two benchmark datasets for zero-shot image classification. The results are compelling to other state-of-the-art, especially on fine-grained dataset.
Zero-Shot Image Classification via Coupled Discriminative Dictionary Learning
2017-01-01
10 pages
Article/Chapter (Book)
Electronic Resource
English
Learning low-rank and discriminative dictionary for image classification
British Library Online Contents | 2014
|Learning Dictionary of Discriminative Part Detectors for Image Categorization and Cosegmentation
British Library Online Contents | 2016
|A Generative/Discriminative Learning Algorithm for Image Classification
British Library Conference Proceedings | 2005
|Image Deblurring with Coupled Dictionary Learning
British Library Online Contents | 2015
|