Abstract:Accurately and swiftly retrieving from fine-grained images is a critical and challenging task. As the key technology for fine-grained image retrieval, deep metric learning aims to learn a mapping space, where samples exhibit two properties: positive concentration and negative separation, facilitating the measurement of similarities between samples. Unsupervised deep metric learning, which obviates the need for labels during training, has garnered widespread attention compared to its supervised counterparts due to its convenience. Current methods in unsupervised deep metric learning face issues such as imbalance in sample construction, difficulty in sample differentiation, and neglect of intrinsic image features. To address these challenges, we propose Manifold and Patch-based Unsupervised Deep Metric Learning (MPUDML) for Fine-Grained Image Retrieval. Specifically, we adopt a manifold similarity-based balanced sampling strategy for constructing more balanced mini-batch samples. Moreover, we leverage soft supervision information obtained from the manifold and cosine similarities between unlabeled images for sample differentiation, effectively reducing the impact of noisy samples. Additionally, we utilize the rich feature information between internal image patches through image patch-level clustering and localization tasks to guide the acquisition of a more comprehensive feature embedding representation, thereby enhancing retrieval performance. Our method, MPUDML, was evaluated against various state-of-the-art unsupervised deep metric learning approaches in fine-grained image retrieval and clustering tasks. Experimental findings indicate that our MPUDML method exceeds other advanced methods in recall ( R @ K ) and Normalized Mutual Information (NMI).

Image distance measurement based on manifold learning and manifold high-order approximation

Learning an Image Manifold for Retrieval

A Novel Image Classification Method Based on Manifold Learning and Gaussian Mixture Model

A Novel Image Distance Based on Gabor Feature and Approximated Manifold

Manifold-manifold distance and its application to face recognition with image sets.

Image-based 3D Model Retrieval Using Manifold Learning

Manifold-Manifold Distance with application to face recognition based on image set

Learning High-Dimensional Correspondence Via Manifold Learning and Local Approximation

Image Classification with Manifold Learning for Out-of-sample Data

Face Recognition Via Domain Adaptation And Manifold Distance Metric Learning

A New Manifold Distance Measure for Visual Object Categorization

Manifold Preserving: an Intrinsic Approach for Semisupervised Distance Metric Learning

A Regularized Approach for Geodesic-Based Semisupervised Multimanifold Learning

Sub-manifold Distance Based Object Recognition in Clutter

Two-Dimensional Maximum Local Variation Based on Image Euclidean Distance for Face Recognition

Multi-manifold Metric Learning for Face Recognition Based on Image Sets

Nearest Manifold Approach for Face Recognition

Image Retrieval Algorithms Based on Manifold Learning

SSMD: Dimensionality Reduction and Classification of Hyperspectral Images Based on Spatial–Spectral Manifold Distance Metric Learning

Head pose estimation based on manifold embedding and distance metric learning

Manifold and patch-based unsupervised deep metric learning for fine-grained image retrieval