Go to:
Logótipo
Você está em: Start > Publications > View > Automatic Augmentation by Hill Climbing
Map of Premises
Principal
Publication

Automatic Augmentation by Hill Climbing

Title
Automatic Augmentation by Hill Climbing
Type
Article in International Conference Proceedings Book
Year
2019
Authors
Ricardo Cruz
(Author)
Other
The person does not belong to the institution. The person does not belong to the institution. The person does not belong to the institution. View Authenticus page Without ORCID
Joaquim F. Pinto da Costa
(Author)
FCUP
View Personal Page You do not have permissions to view the institutional email. Search for Participant Publications View Authenticus page View ORCID page
Jaime S. Cardoso
(Author)
FEUP
View Personal Page You do not have permissions to view the institutional email. Search for Participant Publications View Authenticus page View ORCID page
Conference proceedings International
Pages: 115-124
28th International Conference on Artificial Neural Networks (ICANN)
Tech Univ Munchen, Klinikum Rechts Isar, Munich, GERMANY, SEP 17-19, 2019
Other information
Authenticus ID: P-00R-1ZW
Abstract (EN): When learning from images, it is desirable to augment the dataset with plausible transformations of its images. Unfortunately, it is not always intuitive for the user how much shear or translation to apply. For this reason, training multiple models through hyperparameter search is required to find the best augmentation policies. But these methods are computationally expensive. Furthermore, since they generate static policies, they do not take advantage of smoothly introducing more aggressive augmentation transformations. In this work, we propose repeating each epoch twice with a small difference in data augmentation intensity, walking towards the best policy. This process doubles the number of epochs, but avoids having to train multiple models. The method is compared against random and Bayesian search for classification and segmentation tasks. The proposal improved twice over random search and was on par with Bayesian search for 4% of the training epochs.
Language: English
Type (Professor's evaluation): Scientific
No. of pages: 10
Documents
We could not find any documents associated to the publication.
Related Publications

Of the same authors

Binary ranking for ordinal class imbalance (2018)
Article in International Scientific Journal
Ricardo Cruz; Kelwin Fernandes; Joaquim F. Pinto Costa; María Pérez Ortiz; Jaime S. Cardoso
Averse Deep Semantic Segmentation (2019)
Article in International Conference Proceedings Book
Ricardo Cruz; Joaquim F. Pinto Costa; Jaime S. Cardoso
Recommend this page Top
Copyright 1996-2025 © Faculdade de Medicina Dentária da Universidade do Porto  I Terms and Conditions  I Acessibility  I Index A-Z
Page created on: 2025-07-14 at 13:15:01 | Privacy Policy | Personal Data Protection Policy | Whistleblowing | Electronic Yellow Book