Learning invariant features by harnessing the aperture problem

Roland Memisevic; Georgios Exarchakis

Learning invariant features by harnessing the aperture problem

Roland Memisevic, Georgios Exarchakis ;

Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):100-108, 2013.

Abstract

The energy model is a simple, biologically inspired approach to extracting relationships between images in tasks like stereopsis and motion analysis. We discuss how adding an extra pooling layer to the energy model makes it possible to learn encodings of transformations that are mostly invariant with respect to image content, and to learn encodings of images that are mostly invariant with respect to the observed transformations. We show how this makes it possible to learn 3D pose-invariant features of objects by watching videos of the objects. We test our approach on a dataset of videos derived from the NORB dataset.

Related Material

Download PDF

@InProceedings{pmlr-v28-memisevic13, title = {Learning invariant features by harnessing the aperture problem}, author = {Roland Memisevic and Georgios Exarchakis}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {100--108}, year = {2013}, editor = {Sanjoy Dasgupta and David McAllester}, volume = {28}, number = {3}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/memisevic13.pdf}, url = {http://proceedings.mlr.press/v28/memisevic13.html}, abstract = {The energy model is a simple, biologically inspired approach to extracting relationships between images in tasks like stereopsis and motion analysis. We discuss how adding an extra pooling layer to the energy model makes it possible to learn encodings of transformations that are mostly invariant with respect to image content, and to learn encodings of images that are mostly invariant with respect to the observed transformations. We show how this makes it possible to learn 3D pose-invariant features of objects by watching videos of the objects. We test our approach on a dataset of videos derived from the NORB dataset. } }

%0 Conference Paper %T Learning invariant features by harnessing the aperture problem %A Roland Memisevic %A Georgios Exarchakis %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-memisevic13 %I PMLR %J Proceedings of Machine Learning Research %P 100--108 %U http://proceedings.mlr.press %V 28 %N 3 %W PMLR %X The energy model is a simple, biologically inspired approach to extracting relationships between images in tasks like stereopsis and motion analysis. We discuss how adding an extra pooling layer to the energy model makes it possible to learn encodings of transformations that are mostly invariant with respect to image content, and to learn encodings of images that are mostly invariant with respect to the observed transformations. We show how this makes it possible to learn 3D pose-invariant features of objects by watching videos of the objects. We test our approach on a dataset of videos derived from the NORB dataset.

TY - CPAPER TI - Learning invariant features by harnessing the aperture problem AU - Roland Memisevic AU - Georgios Exarchakis BT - Proceedings of the 30th International Conference on Machine Learning PY - 2013/02/13 DA - 2013/02/13 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-memisevic13 PB - PMLR SP - 100 DP - PMLR EP - 108 L1 - http://proceedings.mlr.press/v28/memisevic13.pdf UR - http://proceedings.mlr.press/v28/memisevic13.html AB - The energy model is a simple, biologically inspired approach to extracting relationships between images in tasks like stereopsis and motion analysis. We discuss how adding an extra pooling layer to the energy model makes it possible to learn encodings of transformations that are mostly invariant with respect to image content, and to learn encodings of images that are mostly invariant with respect to the observed transformations. We show how this makes it possible to learn 3D pose-invariant features of objects by watching videos of the objects. We test our approach on a dataset of videos derived from the NORB dataset. ER -