From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process. Le Cacheux, Y., Le Borgne, H., & Crucianu, M. In International Conference on MultiMedia Modeling (MMM), Thessaloniki, Grece, 1, 2019. ![From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Pdf doi abstract bibtex Zero-shot learning (ZSL) is concerned with the recognition of previously unseen classes. It relies on additional semantic knowledge for which a mapping can be learned with training examples of seen classes. While classical ZSL considers the recognition performance on unseen classes only, generalized zero-shot learning (GZSL) aims at maximizing performance on both seen and unseen classes. In this paper, we propose a new process for training and evaluation in the GZSL setting; this process addresses the gap in performance between samples from unseen and seen classes by penalizing the latter, and enables to select hyper-parameters well-suited to the GZSL task. It can be applied to any existing ZSL approach and leads to a significant performance boost: the experimental evaluation shows that GZSL performance, averaged over eight state-of-the-art methods, is improved from 28.5 to 42.2 on CUB and from 28.2 to 57.1 on AwA2.
@inproceedings{lecacheux2019mmm,
author = {Le Cacheux, Yannick and Le Borgne, Herv{\'e} and Crucianu, Michel},
title = {From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process},
booktitle = {International Conference on MultiMedia Modeling (MMM)},
year = {2019},
address = {Thessaloniki, Grece},
month = {1},
url_PDF = {https://arxiv.org/pdf/1809.10120.pdf},
doi = {10.1007/978-3-030-05716-9_38},
abstract = {Zero-shot learning (ZSL) is concerned with the recognition of previously unseen classes. It relies on additional semantic knowledge for which a mapping can be learned with training examples of seen classes. While classical ZSL considers the recognition performance on unseen classes only, generalized zero-shot learning (GZSL) aims at maximizing performance on both seen and unseen classes. In this paper, we propose a new process for training and evaluation in the GZSL setting; this process addresses the gap in performance between samples from unseen and seen classes by penalizing the latter, and enables to select hyper-parameters well-suited to the GZSL task. It can be applied to any existing ZSL approach and leads to a significant performance boost: the experimental evaluation shows that GZSL performance, averaged over eight state-of-the-art methods, is improved from 28.5 to 42.2 on CUB and from 28.2 to 57.1 on AwA2.},
keywords = {zero-shot-learning}
}
Downloads: 0
{"_id":"qu9vJoP9XhHX8sX8g","bibbaseid":"lecacheux-leborgne-crucianu-fromclassicaltogeneralizedzeroshotlearningasimpleadaptationprocess-2019","author_short":["Le Cacheux, Y.","Le Borgne, H.","Crucianu, M."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"propositions":[],"lastnames":["Le","Cacheux"],"firstnames":["Yannick"],"suffixes":[]},{"propositions":[],"lastnames":["Le","Borgne"],"firstnames":["Hervé"],"suffixes":[]},{"propositions":[],"lastnames":["Crucianu"],"firstnames":["Michel"],"suffixes":[]}],"title":"From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process","booktitle":"International Conference on MultiMedia Modeling (MMM)","year":"2019","address":"Thessaloniki, Grece","month":"1","url_pdf":"https://arxiv.org/pdf/1809.10120.pdf","doi":"10.1007/978-3-030-05716-9_38","abstract":"Zero-shot learning (ZSL) is concerned with the recognition of previously unseen classes. It relies on additional semantic knowledge for which a mapping can be learned with training examples of seen classes. While classical ZSL considers the recognition performance on unseen classes only, generalized zero-shot learning (GZSL) aims at maximizing performance on both seen and unseen classes. In this paper, we propose a new process for training and evaluation in the GZSL setting; this process addresses the gap in performance between samples from unseen and seen classes by penalizing the latter, and enables to select hyper-parameters well-suited to the GZSL task. It can be applied to any existing ZSL approach and leads to a significant performance boost: the experimental evaluation shows that GZSL performance, averaged over eight state-of-the-art methods, is improved from 28.5 to 42.2 on CUB and from 28.2 to 57.1 on AwA2.","keywords":"zero-shot-learning","bibtex":"@inproceedings{lecacheux2019mmm,\n author = {Le Cacheux, Yannick and Le Borgne, Herv{\\'e} and Crucianu, Michel},\n title = {From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process},\n booktitle = {International Conference on MultiMedia Modeling (MMM)},\n year = {2019},\n address = {Thessaloniki, Grece},\n month = {1},\n url_PDF = {https://arxiv.org/pdf/1809.10120.pdf},\n doi = {10.1007/978-3-030-05716-9_38},\n abstract = {Zero-shot learning (ZSL) is concerned with the recognition of previously unseen classes. It relies on additional semantic knowledge for which a mapping can be learned with training examples of seen classes. While classical ZSL considers the recognition performance on unseen classes only, generalized zero-shot learning (GZSL) aims at maximizing performance on both seen and unseen classes. In this paper, we propose a new process for training and evaluation in the GZSL setting; this process addresses the gap in performance between samples from unseen and seen classes by penalizing the latter, and enables to select hyper-parameters well-suited to the GZSL task. It can be applied to any existing ZSL approach and leads to a significant performance boost: the experimental evaluation shows that GZSL performance, averaged over eight state-of-the-art methods, is improved from 28.5 to 42.2 on CUB and from 28.2 to 57.1 on AwA2.},\n keywords = {zero-shot-learning}\n}\n\n","author_short":["Le Cacheux, Y.","Le Borgne, H.","Crucianu, M."],"key":"lecacheux2019mmm","id":"lecacheux2019mmm","bibbaseid":"lecacheux-leborgne-crucianu-fromclassicaltogeneralizedzeroshotlearningasimpleadaptationprocess-2019","role":"author","urls":{" pdf":"https://arxiv.org/pdf/1809.10120.pdf"},"keyword":["zero-shot-learning"],"metadata":{"authorlinks":{}},"html":""},"bibtype":"inproceedings","biburl":"https://hleborgne.github.io/files/hleborgne-publications.bib","dataSources":["sJzmxoNKfHCgQoayi"],"keywords":["zero-shot-learning"],"search_terms":["classical","generalized","zero","shot","learning","simple","adaptation","process","le cacheux","le borgne","crucianu"],"title":"From Classical to Generalized Zero-Shot Learning: a Simple Adaptation Process","year":2019}