Convolutional Low-Resolution Fine-Grained Classification

TL;DR

This paper introduces a novel end-to-end deep model that combines super-resolution and fine-grained classification to improve accuracy on low-resolution images, demonstrating superior performance on benchmark datasets.

Contribution

The paper proposes a resolution-aware CNN that jointly learns super-resolution and fine-grained classification in a unified framework, addressing the challenge of missing details in low-resolution images.

Findings

The model outperforms traditional CNNs on low-resolution fine-grained classification tasks.

Experiments on Stanford Cars and Caltech-UCSD Birds datasets show consistent accuracy improvements.

The approach effectively recovers details lost in low-resolution images for better classification.

Abstract

Successful fine-grained image classification methods learn subtle details between visually similar (sub-)classes, but the problem becomes significantly more challenging if the details are missing due to low resolution. Encouraged by the recent success of Convolutional Neural Network (CNN) architectures in image classification, we propose a novel resolution-aware deep model which combines convolutional image super-resolution and convolutional fine-grained classification into a single model in an end-to-end manner. Extensive experiments on the Stanford Cars and Caltech-UCSD Birds 200-2011 benchmarks demonstrate that the proposed model consistently performs better than conventional convolutional net on classifying fine-grained object classes in low-resolution images.