|
Enric Marti, Debora Gil, Marc Vivet, & Carme Julia. (2009). Aprendizaje Basado en Proyectos en la asignatura de Gráficos por Computador en Ingeniería Informática. Balance de cuatro años de experiencia. Barcelona, Spain.
|
|
|
Miquel Ferrer, Robert Benavente, Ernest Valveny, J. Garcia, Agata Lapedriza, & Gemma Sanchez. (2008). Aprendizaje Cooperativo Aplicado a la Docencia de las Asignaturas de Programacion en Ingenieria Informatica.
|
|
|
Agata Lapedriza, David Masip, & Jordi Vitria. (2005). Are external face features useful for automatic face classification?.
|
|
|
S. Garcia, Dani Rowe, Jordi Gonzalez, & Juan J. Villanueva. (2005). Articulated Object Modelling Using Neural Gas Networks.
|
|
|
Juan J. Villanueva, Jordi Gonzalez, Javier Varona, & Xavier Roca. (2002). Aspaces: Action Spaces for Recognition and Synthesis of Human Actions..
|
|
|
A. Martinez, Jordi Vitria, & S. Sampayo. (1995). Atlas: a Hexapod driven by a Neural Network..
|
|
|
Marcin Przewiezlikowski, Mateusz Pyla, Bartosz Zielinski, Bartłomiej Twardowski, Jacek Tabor, & Marek Smieja. (2023). Augmentation-aware Self-supervised Learning with Guided Projector.
Abstract: Self-supervised learning (SSL) is a powerful technique for learning robust representations from unlabeled data. By learning to remain invariant to applied data augmentations, methods such as SimCLR and MoCo are able to reach quality on par with supervised approaches. However, this invariance may be harmful to solving some downstream tasks which depend on traits affected by augmentations used during pretraining, such as color. In this paper, we propose to foster sensitivity to such characteristics in the representation space by modifying the projector network, a common component of self-supervised architectures. Specifically, we supplement the projector with information about augmentations applied to images. In order for the projector to take advantage of this auxiliary conditioning when solving the SSL task, the feature extractor learns to preserve the augmentation information in its representations. Our approach, coined Conditional Augmentation-aware Self-supervised Learning (CASSLE), is directly applicable to typical joint-embedding SSL methods regardless of their objective functions. Moreover, it does not require major changes in the network architecture or prior knowledge of downstream tasks. In addition to an analysis of sensitivity towards different data augmentations, we conduct a series of experiments, which show that CASSLE improves over various SSL methods, reaching state-of-the-art performance in multiple downstream tasks.
|
|
|
Joan Mas, B. Lamiroy, Gemma Sanchez, & Josep Llados. (2006). Automatic Adjacency Grammar Generation from User Drawn Sketches.
|
|
|
J.M. Sanchez, & X. Binefa. (1999). Automatic digital TV commercial recognition..
|
|
|
Angel Sappa. (2004). Automatic Extraction of Planar Projections from Panoramic Range Images.
|
|
|
Ernest Valveny, & B. Lamiroy. (2002). Automatic Generation of Browsable Technical Documents..
|
|
|
Jose Antonio Rodriguez, Gemma Sanchez, & Josep Llados. (2006). Automatic Interpretation of Proofreading Sketches.
|
|
|
Ellen J.L. Brunenberg, Oriol Pujol, Bart M. Ter Haar Romeny, & Petia Radeva. (2006). Automatic IVUS Segmentation of Atherosclerotic Plaque with Stop & Go Snake.
|
|
|
Jordi Gonzalez, Javier Varona, Xavier Roca, & Juan J. Villanueva. (2003). Automatic Keyframing of Human Actions for Computer Animation.
|
|
|
Jordi Gonzalez, Javier Varona, Xavier Roca, & Juan J. Villanueva. (2003). Automatic Keyframing of Human Actions for Computer Animation.
|
|