Learning to compare: Relation network for few-shot learning
We present a conceptually simple, flexible, and general framework for few-shot learning, where a classifier must learn to recognise new classes given only few examples from each. Our method, called the Relation Network (RN), is trained end-to-end from scratch. During meta-learning, it learns to lear...
| Main Authors: | , , , , , |
|---|---|
| Format: | Conference item |
| Published: |
Institute of Electrical and Electronics Engineers
2018
|
| _version_ | 1797083201301118976 |
|---|---|
| author | Sung, F Yang, Y Zhang, L Xiang, T Torr, P Hospedales, T |
| author_facet | Sung, F Yang, Y Zhang, L Xiang, T Torr, P Hospedales, T |
| author_sort | Sung, F |
| collection | OXFORD |
| description | We present a conceptually simple, flexible, and general framework for few-shot learning, where a classifier must learn to recognise new classes given only few examples from each. Our method, called the Relation Network (RN), is trained end-to-end from scratch. During meta-learning, it learns to learn a deep distance metric to compare a small number of images within episodes, each of which is designed to simulate the few-shot setting. Once trained, a RN is able to classify images of new classes by computing relation scores between query images and the few examples of each new class without further updating the network. Besides providing improved performance on few-shot learning, our framework is easily extended to zero-shot learning. Extensive experiments on five benchmarks demonstrate that our simple approach provides a unified and effective approach for both of these two tasks. |
| first_indexed | 2024-03-07T01:38:21Z |
| format | Conference item |
| id | oxford-uuid:95f7e4a6-5e2d-4349-b0ed-9fb1c173fa4f |
| institution | University of Oxford |
| last_indexed | 2024-03-07T01:38:21Z |
| publishDate | 2018 |
| publisher | Institute of Electrical and Electronics Engineers |
| record_format | dspace |
| spelling | oxford-uuid:95f7e4a6-5e2d-4349-b0ed-9fb1c173fa4f2022-03-26T23:49:54ZLearning to compare: Relation network for few-shot learningConference itemhttp://purl.org/coar/resource_type/c_5794uuid:95f7e4a6-5e2d-4349-b0ed-9fb1c173fa4fSymplectic Elements at OxfordInstitute of Electrical and Electronics Engineers2018Sung, FYang, YZhang, LXiang, TTorr, PHospedales, TWe present a conceptually simple, flexible, and general framework for few-shot learning, where a classifier must learn to recognise new classes given only few examples from each. Our method, called the Relation Network (RN), is trained end-to-end from scratch. During meta-learning, it learns to learn a deep distance metric to compare a small number of images within episodes, each of which is designed to simulate the few-shot setting. Once trained, a RN is able to classify images of new classes by computing relation scores between query images and the few examples of each new class without further updating the network. Besides providing improved performance on few-shot learning, our framework is easily extended to zero-shot learning. Extensive experiments on five benchmarks demonstrate that our simple approach provides a unified and effective approach for both of these two tasks. |
| spellingShingle | Sung, F Yang, Y Zhang, L Xiang, T Torr, P Hospedales, T Learning to compare: Relation network for few-shot learning |
| title | Learning to compare: Relation network for few-shot learning |
| title_full | Learning to compare: Relation network for few-shot learning |
| title_fullStr | Learning to compare: Relation network for few-shot learning |
| title_full_unstemmed | Learning to compare: Relation network for few-shot learning |
| title_short | Learning to compare: Relation network for few-shot learning |
| title_sort | learning to compare relation network for few shot learning |
| work_keys_str_mv | AT sungf learningtocomparerelationnetworkforfewshotlearning AT yangy learningtocomparerelationnetworkforfewshotlearning AT zhangl learningtocomparerelationnetworkforfewshotlearning AT xiangt learningtocomparerelationnetworkforfewshotlearning AT torrp learningtocomparerelationnetworkforfewshotlearning AT hospedalest learningtocomparerelationnetworkforfewshotlearning |