Deep Feature Learning with Manifold Embedding for Robust Image Retrieval
Conventionally, the similarity between two images is measured by the easy-calculating Euclidean distance between their corresponding image feature representations for image retrieval. However, this kind of direct similarity measurement ignores the local geometry structure of the intrinsic data manif...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-12-01
|
| Series: | Algorithms |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1999-4893/13/12/318 |
| _version_ | 1827700944867950592 |
|---|---|
| author | Xin Chen Ying Li |
| author_facet | Xin Chen Ying Li |
| author_sort | Xin Chen |
| collection | DOAJ |
| description | Conventionally, the similarity between two images is measured by the easy-calculating Euclidean distance between their corresponding image feature representations for image retrieval. However, this kind of direct similarity measurement ignores the local geometry structure of the intrinsic data manifold, which is not discriminative enough for robust image retrieval. Some works have proposed to tackle this problem by re-ranking with manifold learning. While benefiting better performance, algorithms of this category suffer from non-trivial computational complexity, which is unfavorable for its application to large-scale retrieval tasks. To address the above problems, in this paper, we propose to learn a robust feature embedding with the guidance of manifold relationships. Specifically, the manifold relationship is used to guide the automatic selection of training image pairs. A fine-tuning network with those selected image pairs transfers such manifold relationships into the fine-tuned feature embedding. With the fine-tuned feature embedding, the Euclidean distance can be directly used to measure the pairwise similarity between images, where the manifold structure is implicitly embedded. Thus, we maintain both the efficiency of Euclidean distance-based similarity measurement and the effectiveness of manifold information in the new feature embedding. Extensive experiments on three benchmark datasets demonstrate the robustness of our proposed method, where our approach significantly outperforms the baselines and exceeds or is comparable to the state-of-the-art methods. |
| first_indexed | 2024-03-10T14:22:55Z |
| format | Article |
| id | doaj.art-ff785b5574664bf9b87b2a6950d85f61 |
| institution | Directory Open Access Journal |
| issn | 1999-4893 |
| language | English |
| last_indexed | 2024-03-10T14:22:55Z |
| publishDate | 2020-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Algorithms |
| spelling | doaj.art-ff785b5574664bf9b87b2a6950d85f612023-11-20T23:15:23ZengMDPI AGAlgorithms1999-48932020-12-01131231810.3390/a13120318Deep Feature Learning with Manifold Embedding for Robust Image RetrievalXin Chen0Ying Li1College of Electronics and Information Engineering, Tongji University, Shanghai 201804, ChinaSchool of Computer and Electronic Information, Nanjing Normal University, Nanjing 210023, ChinaConventionally, the similarity between two images is measured by the easy-calculating Euclidean distance between their corresponding image feature representations for image retrieval. However, this kind of direct similarity measurement ignores the local geometry structure of the intrinsic data manifold, which is not discriminative enough for robust image retrieval. Some works have proposed to tackle this problem by re-ranking with manifold learning. While benefiting better performance, algorithms of this category suffer from non-trivial computational complexity, which is unfavorable for its application to large-scale retrieval tasks. To address the above problems, in this paper, we propose to learn a robust feature embedding with the guidance of manifold relationships. Specifically, the manifold relationship is used to guide the automatic selection of training image pairs. A fine-tuning network with those selected image pairs transfers such manifold relationships into the fine-tuned feature embedding. With the fine-tuned feature embedding, the Euclidean distance can be directly used to measure the pairwise similarity between images, where the manifold structure is implicitly embedded. Thus, we maintain both the efficiency of Euclidean distance-based similarity measurement and the effectiveness of manifold information in the new feature embedding. Extensive experiments on three benchmark datasets demonstrate the robustness of our proposed method, where our approach significantly outperforms the baselines and exceeds or is comparable to the state-of-the-art methods.https://www.mdpi.com/1999-4893/13/12/318image retrievaldeep feature learningsimilarity measurementmanifold embedding |
| spellingShingle | Xin Chen Ying Li Deep Feature Learning with Manifold Embedding for Robust Image Retrieval Algorithms image retrieval deep feature learning similarity measurement manifold embedding |
| title | Deep Feature Learning with Manifold Embedding for Robust Image Retrieval |
| title_full | Deep Feature Learning with Manifold Embedding for Robust Image Retrieval |
| title_fullStr | Deep Feature Learning with Manifold Embedding for Robust Image Retrieval |
| title_full_unstemmed | Deep Feature Learning with Manifold Embedding for Robust Image Retrieval |
| title_short | Deep Feature Learning with Manifold Embedding for Robust Image Retrieval |
| title_sort | deep feature learning with manifold embedding for robust image retrieval |
| topic | image retrieval deep feature learning similarity measurement manifold embedding |
| url | https://www.mdpi.com/1999-4893/13/12/318 |
| work_keys_str_mv | AT xinchen deepfeaturelearningwithmanifoldembeddingforrobustimageretrieval AT yingli deepfeaturelearningwithmanifoldembeddingforrobustimageretrieval |