Object retrieval with large vocabularies and fast spatial matching
In this paper, we present a large-scale object retrieval system. The user supplies a query object by selecting a region of a query image, and the system returns a ranked list of images that contain the same object, retrieved from a large corpus. We demonstrate the scalability and performance of our...
| Main Authors: | , , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
IEEE
2007
|
| _version_ | 1824459115660836864 |
|---|---|
| author | Philbin, J Chum, O Isard, M Sivic, J Zisserman, A |
| author_facet | Philbin, J Chum, O Isard, M Sivic, J Zisserman, A |
| author_sort | Philbin, J |
| collection | OXFORD |
| description | In this paper, we present a large-scale object retrieval system. The user supplies a query object by selecting a region of a query image, and the system returns a ranked list of images that contain the same object, retrieved from a large corpus. We demonstrate the scalability and performance of our system on a dataset of over 1 million images crawled from the photo-sharing site, Flickr [3], using Oxford landmarks as queries. Building an image-feature vocabulary is a major time and performance bottleneck, due to the size of our dataset. To address this problem we compare different scalable methods for building a vocabulary and introduce a novel quantization method based on randomized trees which we show outperforms the current state-of-the-art on an extensive ground-truth. Our experiments show that the quantization has a major effect on retrieval quality. To further improve query performance, we add an efficient spatial verification stage to re-rank the results returned from our bag-of-words model and show that this consistently improves search quality, though by less of a margin when the visual vocabulary is large. We view this work as a promising step towards much larger, "web-scale" image corpora. |
| first_indexed | 2025-02-19T04:36:39Z |
| format | Conference item |
| id | oxford-uuid:a2e03ceb-979d-4dbe-af06-0511621a1ca6 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2025-02-19T04:36:39Z |
| publishDate | 2007 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:a2e03ceb-979d-4dbe-af06-0511621a1ca62025-01-31T11:37:45ZObject retrieval with large vocabularies and fast spatial matchingConference itemhttp://purl.org/coar/resource_type/c_5794uuid:a2e03ceb-979d-4dbe-af06-0511621a1ca6EnglishSymplectic ElementsIEEE2007Philbin, JChum, OIsard, MSivic, JZisserman, AIn this paper, we present a large-scale object retrieval system. The user supplies a query object by selecting a region of a query image, and the system returns a ranked list of images that contain the same object, retrieved from a large corpus. We demonstrate the scalability and performance of our system on a dataset of over 1 million images crawled from the photo-sharing site, Flickr [3], using Oxford landmarks as queries. Building an image-feature vocabulary is a major time and performance bottleneck, due to the size of our dataset. To address this problem we compare different scalable methods for building a vocabulary and introduce a novel quantization method based on randomized trees which we show outperforms the current state-of-the-art on an extensive ground-truth. Our experiments show that the quantization has a major effect on retrieval quality. To further improve query performance, we add an efficient spatial verification stage to re-rank the results returned from our bag-of-words model and show that this consistently improves search quality, though by less of a margin when the visual vocabulary is large. We view this work as a promising step towards much larger, "web-scale" image corpora. |
| spellingShingle | Philbin, J Chum, O Isard, M Sivic, J Zisserman, A Object retrieval with large vocabularies and fast spatial matching |
| title | Object retrieval with large vocabularies and fast spatial matching |
| title_full | Object retrieval with large vocabularies and fast spatial matching |
| title_fullStr | Object retrieval with large vocabularies and fast spatial matching |
| title_full_unstemmed | Object retrieval with large vocabularies and fast spatial matching |
| title_short | Object retrieval with large vocabularies and fast spatial matching |
| title_sort | object retrieval with large vocabularies and fast spatial matching |
| work_keys_str_mv | AT philbinj objectretrievalwithlargevocabulariesandfastspatialmatching AT chumo objectretrievalwithlargevocabulariesandfastspatialmatching AT isardm objectretrievalwithlargevocabulariesandfastspatialmatching AT sivicj objectretrievalwithlargevocabulariesandfastspatialmatching AT zissermana objectretrievalwithlargevocabulariesandfastspatialmatching |