FineTag: Multi-attribute Classification at Fine-grained Level in Images

TL;DR

This paper introduces FineTag, an end-to-end bi-linear CNN architecture with pairwise ranking loss for fine-grained multi-attribute classification, outperforming baseline models with fewer parameters on the CUB200 bird dataset.

Contribution

First application of bi-linear CNN with pairwise ranking loss for fine-grained attribute classification, achieving high performance with fewer parameters.

Findings

Outperforms baseline CNN models in accuracy.

Uses 40 times fewer parameters than competitors.

Validated on CUB200 bird dataset with adapted annotations.

Abstract

In this paper, we address the extraction of the fine-grained attributes of an instance as a `multi-attribute classification' problem. To this end, we propose an end-to-end architecture by adopting the bi-linear Convolutional Neural Network with the pairwise ranking loss. This is the first time such architecture is applied for the fine-grained attributes classification problem. We compared the proposed method with a competitive deep Convolutional Neural Network baseline. Extensive experiments show that the proposed method attains/outperforms the performance of compared baseline with significantly less number of parameters ($40\times$ less). We demonstrated our approach on CUB200 birds dataset whose annotations are adapted in this work for multi-attribute classification at fine-grained level.