Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition
Citations
6,150 citations
Cites background or methods from "Synthetic Data and Artificial Neura..."
...In this work we introduce a new learnable module, the Spatial Transformer, which explicitly allows the spatial manipulation of data within the network....
[...]
...5....
[...]
...While we only explore feed-forward networks in this work, early experiments show spatial transformers to be powerful in recurrent models, and useful for tasks requiring the disentangling of object reference frames, as well as easily extendable to 3D transformations (see Appendix A.3)....
[...]
5,782 citations
Cites background from "Synthetic Data and Artificial Neura..."
...[124] train exclusively with synthetic data for natural scene text recognition....
[...]
2,184 citations
1,159 citations
Cites background from "Synthetic Data and Artificial Neura..."
...The combination of an endto-end learning system with minimal need for human design decisions, and the ability to efficiently train large and complex models, have allowed them to achieve state-of-the-art performance in a number of benchmarks [10, 14, 19, 33, 37, 38]....
[...]
1,142 citations
References
73,978 citations
42,067 citations
6,899 citations
"Synthetic Data and Artificial Neura..." refers methods in this paper
...We train the network by back-propagating the standard multinomial logistic regression loss with dropout [10], which improves generalization....
[...]
3,472 citations
"Synthetic Data and Artificial Neura..." refers methods in this paper
...ter sequence encoding (Sect. 3.2), and bag-of-N-grams encoding (Sect. 3.3). Our recognition methods work by performing inference on the word image holistically, mimicking more closely how humans read [6], rather than by sequential character classification as is traditionally done. Due to the large training data requirements of these models, we also present a synthetic data generation method. This synt...
[...]
3,043 citations