DRAW: A Recurrent Neural Network For Image Generation

Karol Gregor, Ivo Danihelka, Alex Graves, Danilo Rezende, Daan Wierstra
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:1462-1471, 2015.

Abstract

This paper introduces the Deep Recurrent Attentive Writer (DRAW) architecture for image generation with neural networks. DRAW networks combine a novel spatial attention mechanism that mimics the foveation of the human eye, with a sequential variational auto-encoding framework that allows for the iterative construction of complex images. The system substantially improves on the state of the art for generative models on MNIST, and, when trained on the Street View House Numbers dataset, it is able to generate images that are indistinguishable from real data with the naked eye.

Cite this Paper

BibTeX
@InProceedings{pmlr-v37-gregor15, title = {DRAW: A Recurrent Neural Network For Image Generation}, author = {Gregor, Karol and Danihelka, Ivo and Graves, Alex and Rezende, Danilo and Wierstra, Daan}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {1462--1471}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/gregor15.pdf}, url = {https://proceedings.mlr.press/v37/gregor15.html}, abstract = {This paper introduces the Deep Recurrent Attentive Writer (DRAW) architecture for image generation with neural networks. DRAW networks combine a novel spatial attention mechanism that mimics the foveation of the human eye, with a sequential variational auto-encoding framework that allows for the iterative construction of complex images. The system substantially improves on the state of the art for generative models on MNIST, and, when trained on the Street View House Numbers dataset, it is able to generate images that are indistinguishable from real data with the naked eye.} }
Endnote
%0 Conference Paper %T DRAW: A Recurrent Neural Network For Image Generation %A Karol Gregor %A Ivo Danihelka %A Alex Graves %A Danilo Rezende %A Daan Wierstra %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-gregor15 %I PMLR %P 1462--1471 %U https://proceedings.mlr.press/v37/gregor15.html %V 37 %X This paper introduces the Deep Recurrent Attentive Writer (DRAW) architecture for image generation with neural networks. DRAW networks combine a novel spatial attention mechanism that mimics the foveation of the human eye, with a sequential variational auto-encoding framework that allows for the iterative construction of complex images. The system substantially improves on the state of the art for generative models on MNIST, and, when trained on the Street View House Numbers dataset, it is able to generate images that are indistinguishable from real data with the naked eye.
RIS
TY - CPAPER TI - DRAW: A Recurrent Neural Network For Image Generation AU - Karol Gregor AU - Ivo Danihelka AU - Alex Graves AU - Danilo Rezende AU - Daan Wierstra BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-gregor15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 1462 EP - 1471 L1 - http://proceedings.mlr.press/v37/gregor15.pdf UR - https://proceedings.mlr.press/v37/gregor15.html AB - This paper introduces the Deep Recurrent Attentive Writer (DRAW) architecture for image generation with neural networks. DRAW networks combine a novel spatial attention mechanism that mimics the foveation of the human eye, with a sequential variational auto-encoding framework that allows for the iterative construction of complex images. The system substantially improves on the state of the art for generative models on MNIST, and, when trained on the Street View House Numbers dataset, it is able to generate images that are indistinguishable from real data with the naked eye. ER -
APA
Gregor, K., Danihelka, I., Graves, A., Rezende, D. & Wierstra, D.. (2015). DRAW: A Recurrent Neural Network For Image Generation. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:1462-1471 Available from https://proceedings.mlr.press/v37/gregor15.html.

Related Material