-
Real-World Font Recognition Using Deep Network and Domain Adaptation
Authors:
Zhangyang Wang,
Jianchao Yang,
Hailin **,
Eli Shechtman,
Aseem Agarwala,
Jonathan Brandt,
Thomas S. Huang
Abstract:
We address a challenging fine-grain classification problem: recognizing a font style from an image of text. In this task, it is very easy to generate lots of rendered font examples but very hard to obtain real-world labeled images. This real-to-synthetic domain gap caused poor generalization to new real data in previous methods (Chen et al. (2014)). In this paper, we refer to Convolutional Neural…
▽ More
We address a challenging fine-grain classification problem: recognizing a font style from an image of text. In this task, it is very easy to generate lots of rendered font examples but very hard to obtain real-world labeled images. This real-to-synthetic domain gap caused poor generalization to new real data in previous methods (Chen et al. (2014)). In this paper, we refer to Convolutional Neural Networks, and use an adaptation technique based on a Stacked Convolutional Auto-Encoder that exploits unlabeled real-world images combined with synthetic data. The proposed method achieves an accuracy of higher than 80% (top-5) on a real-world dataset.
△ Less
Submitted 31 March, 2015;
originally announced April 2015.
-
Decomposition-Based Domain Adaptation for Real-World Font Recognition
Authors:
Zhangyang Wang,
Jianchao Yang,
Hailin **,
Eli Shechtman,
Aseem Agarwala,
Jonathan Brandt,
Thomas S. Huang
Abstract:
We present a domain adaption framework to address a domain mismatch between synthetic training and real-world testing data. We demonstrate our method on a challenging fine-grain classification problem: recognizing a font style from an image of text. In this task, it is very easy to generate lots of rendered font examples but very hard to obtain real-world labeled images. This real-to-synthetic dom…
▽ More
We present a domain adaption framework to address a domain mismatch between synthetic training and real-world testing data. We demonstrate our method on a challenging fine-grain classification problem: recognizing a font style from an image of text. In this task, it is very easy to generate lots of rendered font examples but very hard to obtain real-world labeled images. This real-to-synthetic domain gap caused poor generalization to new real data in previous font recognition methods (Chen et al. (2014)). In this paper, we introduce a Convolutional Neural Network decomposition approach, leveraging a large training corpus of synthetic data to obtain effective features for classification. This is done using an adaptation technique based on a Stacked Convolutional Auto-Encoder that exploits a large collection of unlabeled real-world text images combined with synthetic data preprocessed in a specific way. The proposed DeepFont method achieves an accuracy of higher than 80% (top-5) on a new large labeled real-world dataset we collected.
△ Less
Submitted 1 April, 2015; v1 submitted 18 December, 2014;
originally announced December 2014.
-
Recognizing Image Style
Authors:
Sergey Karayev,
Matthew Trentacoste,
Helen Han,
Aseem Agarwala,
Trevor Darrell,
Aaron Hertzmann,
Holger Winnemoeller
Abstract:
The style of an image plays a significant role in how it is viewed, but style has received little attention in computer vision research. We describe an approach to predicting style of images, and perform a thorough evaluation of different image features for these tasks. We find that features learned in a multi-layer network generally perform best -- even when trained with object class (not style)…
▽ More
The style of an image plays a significant role in how it is viewed, but style has received little attention in computer vision research. We describe an approach to predicting style of images, and perform a thorough evaluation of different image features for these tasks. We find that features learned in a multi-layer network generally perform best -- even when trained with object class (not style) labels. Our large-scale learning methods results in the best published performance on an existing dataset of aesthetic ratings and photographic style annotations. We present two novel datasets: 80K Flickr photographs annotated with 20 curated style labels, and 85K paintings annotated with 25 style/genre labels. Our approach shows excellent classification performance on both datasets. We use the learned classifiers to extend traditional tag-based image search to consider stylistic constraints, and demonstrate cross-dataset understanding of style.
△ Less
Submitted 23 July, 2014; v1 submitted 14 November, 2013;
originally announced November 2013.
-
Fock space exploration by angle resolved transmission through quantum diffraction grating of cold atoms in an optical lattice
Authors:
Adhip Agarwala,
Madhurima Nath,
Jasleen Lugani,
K. Thyagarajan,
Sankalpa Ghosh
Abstract:
Light transmission or diffraction from different quantum phases of cold atoms in an optical lattice has recently come up as a useful tool to probe such ultra cold atomic systems. The periodic nature of the optical lattice potential closely resembles the structure of a diffraction grating in real space, but loaded with a strongly correlated quantum many body state which interacts with the incident…
▽ More
Light transmission or diffraction from different quantum phases of cold atoms in an optical lattice has recently come up as a useful tool to probe such ultra cold atomic systems. The periodic nature of the optical lattice potential closely resembles the structure of a diffraction grating in real space, but loaded with a strongly correlated quantum many body state which interacts with the incident electromagnetic wave, a feature that controls the nature of the light transmission or dispersion through such quantum medium. In this paper we show that as one varies the relative angle between the cavity mode and the optical lattice, the peak of the transmission spectrum through such cavity also changes reflecting the statistical distribution of the atoms in the illuminated sites. Consequently the angle resolved transmission spectrum of such quantum diffraction grating can provide a plethora of information about the Fock space structure of the many body quantum state of ultra cold atoms in such an optical cavity that can be explored in current state of the art experiments.
△ Less
Submitted 25 May, 2012; v1 submitted 23 February, 2012;
originally announced February 2012.
