A Two-Layer Local Constrained Sparse Coding Method for Fine-Grained Visual Categorization

TL;DR

This paper introduces a two-layer local constrained sparse coding framework for fine-grained visual categorization, improving discriminative feature learning and training efficiency, achieving competitive accuracy on benchmark datasets.

Contribution

It proposes a novel two-layer sparse coding architecture with local constraints and a quick dictionary update for enhanced FGVC performance.

Findings

Achieved 85.29% accuracy on Oxford 102 flowers dataset.

Achieved 67.8% accuracy on CUB-200-2011 bird dataset.

Framework is highly competitive with existing methods.

Abstract

Fine-grained categories are more difficulty distinguished than generic categories due to the similarity of inter-class and the diversity of intra-class. Therefore, the fine-grained visual categorization (FGVC) is considered as one of challenge problems in computer vision recently. A new feature learning framework, which is based on a two-layer local constrained sparse coding architecture, is proposed in this paper. The two-layer architecture is introduced for learning intermediate-level features, and the local constrained term is applied to guarantee the local smooth of coding coefficients. For extracting more discriminative information, local orientation histograms are the input of sparse coding instead of raw pixels. Moreover, a quick dictionary updating process is derived to further improve the training speed. Two experimental results show that our method achieves 85.29% accuracy on the Oxford 102 flowers dataset and 67.8% accuracy on the CUB-200-2011 bird dataset, and the performance of our framework is highly competitive with existing literatures.