Glocal Retriever: Glocal Image Retrieval Using the Combination of Global and Local Descriptors
Development of deep learning has led to progress in computer vision, including metric learning tasks such as image retrieval, through convolutional neural networks. In image retrieval, the metric distance (i.e., the similarity) between the images needs to be computed and then compared to return simi...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-01-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/12/2/442 |
| _version_ | 1797443428005445632 |
|---|---|
| author | Zeu Kim Youngin Kim Young-Joo Suh |
| author_facet | Zeu Kim Youngin Kim Young-Joo Suh |
| author_sort | Zeu Kim |
| collection | DOAJ |
| description | Development of deep learning has led to progress in computer vision, including metric learning tasks such as image retrieval, through convolutional neural networks. In image retrieval, the metric distance (i.e., the similarity) between the images needs to be computed and then compared to return similar images. Global descriptors are good at extracting holistic features of an image, such as the overall shape of the main object and the silhouette. On the other hand, the local features extract the detailed features which the model uses to help classify similar images together. This paper proposes a descriptor mixer which takes advantage of both local and global descriptors (group of features combined into one) as well as different types of global descriptors for an effect of a lighter version of an ensemble of models (i.e., fewer parameters and smaller model size than those of actual ensemble of networks). As a result, the model’s performance improved about 1.36% (recall @ 32) when the combination of the descriptors were used. We empirically found out that the combination of GeM and MAC achieved the highest performance. |
| first_indexed | 2024-03-09T12:55:54Z |
| format | Article |
| id | doaj.art-614afccf285b4b13a5cd7748e6139409 |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-09T12:55:54Z |
| publishDate | 2023-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-614afccf285b4b13a5cd7748e61394092023-11-30T22:00:31ZengMDPI AGElectronics2079-92922023-01-0112244210.3390/electronics12020442Glocal Retriever: Glocal Image Retrieval Using the Combination of Global and Local DescriptorsZeu Kim0Youngin Kim1Young-Joo Suh2Graduate School of Artificial Intelligence, Pohang University of Science and Technology, Pohang 37673, Republic of KoreaGraduate School of Artificial Intelligence, Pohang University of Science and Technology, Pohang 37673, Republic of KoreaGraduate School of Artificial Intelligence, Pohang University of Science and Technology, Pohang 37673, Republic of KoreaDevelopment of deep learning has led to progress in computer vision, including metric learning tasks such as image retrieval, through convolutional neural networks. In image retrieval, the metric distance (i.e., the similarity) between the images needs to be computed and then compared to return similar images. Global descriptors are good at extracting holistic features of an image, such as the overall shape of the main object and the silhouette. On the other hand, the local features extract the detailed features which the model uses to help classify similar images together. This paper proposes a descriptor mixer which takes advantage of both local and global descriptors (group of features combined into one) as well as different types of global descriptors for an effect of a lighter version of an ensemble of models (i.e., fewer parameters and smaller model size than those of actual ensemble of networks). As a result, the model’s performance improved about 1.36% (recall @ 32) when the combination of the descriptors were used. We empirically found out that the combination of GeM and MAC achieved the highest performance.https://www.mdpi.com/2079-9292/12/2/442image retrievaldeep metric learningcomputer vision |
| spellingShingle | Zeu Kim Youngin Kim Young-Joo Suh Glocal Retriever: Glocal Image Retrieval Using the Combination of Global and Local Descriptors Electronics image retrieval deep metric learning computer vision |
| title | Glocal Retriever: Glocal Image Retrieval Using the Combination of Global and Local Descriptors |
| title_full | Glocal Retriever: Glocal Image Retrieval Using the Combination of Global and Local Descriptors |
| title_fullStr | Glocal Retriever: Glocal Image Retrieval Using the Combination of Global and Local Descriptors |
| title_full_unstemmed | Glocal Retriever: Glocal Image Retrieval Using the Combination of Global and Local Descriptors |
| title_short | Glocal Retriever: Glocal Image Retrieval Using the Combination of Global and Local Descriptors |
| title_sort | glocal retriever glocal image retrieval using the combination of global and local descriptors |
| topic | image retrieval deep metric learning computer vision |
| url | https://www.mdpi.com/2079-9292/12/2/442 |
| work_keys_str_mv | AT zeukim glocalretrieverglocalimageretrievalusingthecombinationofglobalandlocaldescriptors AT younginkim glocalretrieverglocalimageretrievalusingthecombinationofglobalandlocaldescriptors AT youngjoosuh glocalretrieverglocalimageretrievalusingthecombinationofglobalandlocaldescriptors |