Image-Text Pre-Training for Logo Recognition

TL;DR

This paper introduces a novel image-text pre-training approach and an improved metric learning loss to enhance open-set logo recognition, significantly boosting performance across multiple datasets.

Contribution

It proposes using image-text paired pre-training and a new loss function, ProxyNCAHN++, to improve logo matching accuracy, especially for text-rich logos.

Findings

Pre-training on image-text data improves logo retrieval performance.

The method achieves state-of-the-art results on five public logo datasets.

Significant performance gains in zero-shot recall@1 across datasets.

Abstract

Open-set logo recognition is commonly solved by first detecting possible logo regions and then matching the detected parts against an ever-evolving dataset of cropped logo images. The matching model, a metric learning problem, is especially challenging for logo recognition due to the mixture of text and symbols in logos. We propose two novel contributions to improve the matching model's performance: (a) using image-text paired samples for pre-training, and (b) an improved metric learning loss function. A standard paradigm of fine-tuning ImageNet pre-trained models fails to discover the text sensitivity necessary to solve the matching problem effectively. This work demonstrates the importance of pre-training on image-text pairs, which significantly improves the performance of a visual embedder trained for the logo retrieval task, especially for more text-dominant classes. We construct a composite public logo dataset combining LogoDet3K, OpenLogo, and FlickrLogos-47 deemed OpenLogoDet3K47. We show that the same vision backbone pre-trained on image-text data, when fine-tuned on OpenLogoDet3K47, achieves $98.6\%$ recall@1, significantly improving performance over pre-training on Imagenet1K ($97.6\%$). We generalize the ProxyNCA++ loss function to propose ProxyNCAHN++ which incorporates class-specific hard negative images. The proposed method sets new state-of-the-art on five public logo datasets considered, with a $3.5\%$ zero-shot recall@1 improvement on LogoDet3K test, $4\%$ on OpenLogo, $6.5\%$ on FlickrLogos-47, $6.2\%$ on Logos In The Wild, and $0.6\%$ on BelgaLogo.