On Learning Discriminative Features from Synthesized Data for Self-Supervised Fine-Grained Visual Recognition

TL;DR

This paper proposes a novel self-supervised learning strategy that synthesizes data to enhance the extraction of discriminative features, significantly improving fine-grained visual recognition performance.

Contribution

It introduces a data synthesis method that guides SSL models to focus on discriminative features by perturbing non-discriminative ones, advancing FGVR capabilities.

Findings

Improved FGVR accuracy across multiple datasets

Effective identification and perturbation of non-discriminative features

Enhanced model invariance to irrelevant feature variations

Abstract

Self-Supervised Learning (SSL) has become a prominent approach for acquiring visual representations across various tasks, yet its application in fine-grained visual recognition (FGVR) is challenged by the intricate task of distinguishing subtle differences between categories. To overcome this, we introduce an novel strategy that boosts SSL's ability to extract critical discriminative features vital for FGVR. This approach creates synthesized data pairs to guide the model to focus on discriminative features critical for FGVR during SSL. We start by identifying non-discriminative features using two main criteria: features with low variance that fail to effectively separate data and those deemed less important by Grad-CAM induced from the SSL loss. We then introduce perturbations to these non-discriminative features while preserving discriminative ones. A decoder is employed to reconstruct images from both perturbed and original feature vectors to create data pairs. An encoder is trained on such generated data pairs to become invariant to variations in non-discriminative dimensions while focusing on discriminative features, thereby improving the model's performance in FGVR tasks. We demonstrate the promising FGVR performance of the proposed approach through extensive evaluation on a wide variety of datasets.