Image Retrieval Based on Learning to Rank and Multiple Loss
Image retrieval applying deep convolutional features has achieved the most advanced performance in most standard benchmark tests. In image retrieval, deep metric learning (DML) plays a key role and aims to capture semantic similarity information carried by data points. However, two factors may imped...
Published in: | ISPRS International Journal of Geo-Information |
---|---|
Main Authors: | , , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
MDPI AG
2019
|
Subjects: | |
Online Access: | https://doi.org/10.3390/ijgi8090393 https://doaj.org/article/245a2f35d37242a599a15c464c642d79 |
id |
fttriple:oai:gotriple.eu:oai:doaj.org/article:245a2f35d37242a599a15c464c642d79 |
---|---|
record_format |
openpolar |
spelling |
fttriple:oai:gotriple.eu:oai:doaj.org/article:245a2f35d37242a599a15c464c642d79 2023-05-15T16:01:59+02:00 Image Retrieval Based on Learning to Rank and Multiple Loss Lili Fan Hongwei Zhao Haoyu Zhao Pingping Liu Huangshui Hu 2019-09-01 https://doi.org/10.3390/ijgi8090393 https://doaj.org/article/245a2f35d37242a599a15c464c642d79 en eng MDPI AG 2220-9964 doi:10.3390/ijgi8090393 https://doaj.org/article/245a2f35d37242a599a15c464c642d79 undefined ISPRS International Journal of Geo-Information, Vol 8, Iss 9, p 393 (2019) multiple loss function computer vision deep image retrieval learning to rank deep learning info stat Journal Article https://vocabularies.coar-repositories.org/resource_types/c_6501/ 2019 fttriple https://doi.org/10.3390/ijgi8090393 2023-01-22T18:14:10Z Image retrieval applying deep convolutional features has achieved the most advanced performance in most standard benchmark tests. In image retrieval, deep metric learning (DML) plays a key role and aims to capture semantic similarity information carried by data points. However, two factors may impede the accuracy of image retrieval. First, when learning the similarity of negative examples, current methods separate negative pairs into equal distance in the embedding space. Thus, the intraclass data distribution might be missed. Second, given a query, either a fraction of data points, or all of them, are incorporated to build up the similarity structure, which makes it rather complex to calculate similarity or to choose example pairs. In this study, in order to achieve more accurate image retrieval, we proposed a method based on learning to rank and multiple loss (LRML). To address the first problem, through learning the ranking sequence, we separate the negative pairs from the query image into different distance. To tackle the second problem, we used a positive example in the gallery and negative sets from the bottom five ranked by similarity, thereby enhancing training efficiency. Our significant experimental results demonstrate that the proposed method achieves state-of-the-art performance on three widely used benchmarks. Article in Journal/Newspaper DML Unknown The Gallery ENVELOPE(-86.417,-86.417,72.535,72.535) ISPRS International Journal of Geo-Information 8 9 393 |
institution |
Open Polar |
collection |
Unknown |
op_collection_id |
fttriple |
language |
English |
topic |
multiple loss function computer vision deep image retrieval learning to rank deep learning info stat |
spellingShingle |
multiple loss function computer vision deep image retrieval learning to rank deep learning info stat Lili Fan Hongwei Zhao Haoyu Zhao Pingping Liu Huangshui Hu Image Retrieval Based on Learning to Rank and Multiple Loss |
topic_facet |
multiple loss function computer vision deep image retrieval learning to rank deep learning info stat |
description |
Image retrieval applying deep convolutional features has achieved the most advanced performance in most standard benchmark tests. In image retrieval, deep metric learning (DML) plays a key role and aims to capture semantic similarity information carried by data points. However, two factors may impede the accuracy of image retrieval. First, when learning the similarity of negative examples, current methods separate negative pairs into equal distance in the embedding space. Thus, the intraclass data distribution might be missed. Second, given a query, either a fraction of data points, or all of them, are incorporated to build up the similarity structure, which makes it rather complex to calculate similarity or to choose example pairs. In this study, in order to achieve more accurate image retrieval, we proposed a method based on learning to rank and multiple loss (LRML). To address the first problem, through learning the ranking sequence, we separate the negative pairs from the query image into different distance. To tackle the second problem, we used a positive example in the gallery and negative sets from the bottom five ranked by similarity, thereby enhancing training efficiency. Our significant experimental results demonstrate that the proposed method achieves state-of-the-art performance on three widely used benchmarks. |
format |
Article in Journal/Newspaper |
author |
Lili Fan Hongwei Zhao Haoyu Zhao Pingping Liu Huangshui Hu |
author_facet |
Lili Fan Hongwei Zhao Haoyu Zhao Pingping Liu Huangshui Hu |
author_sort |
Lili Fan |
title |
Image Retrieval Based on Learning to Rank and Multiple Loss |
title_short |
Image Retrieval Based on Learning to Rank and Multiple Loss |
title_full |
Image Retrieval Based on Learning to Rank and Multiple Loss |
title_fullStr |
Image Retrieval Based on Learning to Rank and Multiple Loss |
title_full_unstemmed |
Image Retrieval Based on Learning to Rank and Multiple Loss |
title_sort |
image retrieval based on learning to rank and multiple loss |
publisher |
MDPI AG |
publishDate |
2019 |
url |
https://doi.org/10.3390/ijgi8090393 https://doaj.org/article/245a2f35d37242a599a15c464c642d79 |
long_lat |
ENVELOPE(-86.417,-86.417,72.535,72.535) |
geographic |
The Gallery |
geographic_facet |
The Gallery |
genre |
DML |
genre_facet |
DML |
op_source |
ISPRS International Journal of Geo-Information, Vol 8, Iss 9, p 393 (2019) |
op_relation |
2220-9964 doi:10.3390/ijgi8090393 https://doaj.org/article/245a2f35d37242a599a15c464c642d79 |
op_rights |
undefined |
op_doi |
https://doi.org/10.3390/ijgi8090393 |
container_title |
ISPRS International Journal of Geo-Information |
container_volume |
8 |
container_issue |
9 |
container_start_page |
393 |
_version_ |
1766397644045811712 |