![](bg9.png)
containing digits. Thus, the learned model is showed
to be transferable to real handwritten documents,
without any preprocessing.
On unlabeled historical RECITAL documents, our
model detected most of the word/number structures.
However, we showed that the model hardly distin-
guish word from number structures in calligraphic
structures. Also, the model seems to miss few num-
ber structures, mainly due to the confusion between
”1” and ”l” structures.
Despite these limitations, considering the high
variability of the RECITAL documents, in terms of
shape, structure and handwriting style, we observed a
good transferability of our model.
To improve the classification performance of our
model on unlabeled data, future work will focus on
adding more variability of handwriting styles and
structures in the generation of artificial documents.
Then, this classification map will be embedded in a
larger document analysis system.
Source codes for the artificial document generator
and for the structure detection system are available at
https://github.com/GeoTrouvetout/CIRESFI.
REFERENCES
Augustin, E., Brodin, J.-m., Carr
´
e, M., Geoffrois, E.,
Grosicki, E., and Pr
ˆ
eteux, F. (2006). RIMES evalu-
ation campaign for handwritten mail processing. In
Proc. of the Workshop on Frontiers in Handwriting
Recognition, number 1.
Bergstra, J., Breuleux, O., Bastien, F., Lamblin, P., Pascanu,
R., Desjardins, G., Turian, J., Warde-Farley, D., and
Bengio, Y. (2010). Theano: A cpu and gpu math com-
piler in python. In Proc. 9th Python in Science Conf,
pages 1–7.
Butt, U. M., Ahmad, S., Shafait, F., Nansen, C., Mian, A. S.,
and Malik, M. I. (2016). Automatic signature segmen-
tation using hyper-spectral imaging. In Frontiers in
Handwriting Recognition (ICFHR), 2016 15th Inter-
national Conference on, pages 19–24. IEEE.
Cethefi, T. (2016). Project anr-14-ce31-0017 ”contrainte
et int
´
egration : pour une r
´
e
´
evaluation des spectacles
forains et italiens sous l’ancien r
´
egime”.
Delalandre, M., Valveny, E., Pridmore, T., and Karatzas,
D. (2010). Generation of synthetic documents for per-
formance evaluation of symbol recognition & spotting
systems. International journal on document analysis
and recognition, 13(3):187–207.
Dieleman, S., Schl
¨
uter, J., Raffel, C., Olson, E., Sønderby,
S. K., Nouri, D., Maturana, D., Thoma, M., Bat-
tenberg, E., Kelly, J., Fauw, J. D., Heilman, M.,
de Almeida, D. M., McFee, B., Weideman, H.,
Tak
´
acs, G., de Rivaz, P., Crall, J., Sanders, G., Ra-
sul, K., Liu, C., French, G., and Degrave, J. (2015).
Lasagne: First release.
Dumoulin, V. and Visin, F. (2016). A guide to convo-
lution arithmetic for deep learning. arXiv preprint
arXiv:1603.07285.
Gorodkin, J. (2004). Comparing two k-category assign-
ments by a k-category correlation coefficient. Com-
putational biology and chemistry, 28(5):367–374.
Grosicki, E. and El-Abed, H. (2011). ICDAR 2011: French
handwriting recognition competition. In Proc. of IC-
DAR, pages 1459–1463.
Kieu, V. C., Journet, N., Visani, M., Mullot, R., and
Domenger, J.-P. (2013). Semi-synthetic Docu-
ment Image Generation Using Texture Mapping on
Scanned 3D Document Shapes. In The Twelfth In-
ternational Conference on Document Analysis and
Recognition, United States.
Kingma, D. and Ba, J. (2014). Adam: A method
for stochastic optimization. arXiv preprint
arXiv:1412.6980.
Krizhevsky, A., Sutskever, I., and Hinton, G. E. (2012). Im-
agenet classification with deep convolutional neural
networks. In Advances in neural information process-
ing systems, pages 1097–1105.
LeCun, Y., Bengio, Y., and Hinton, G. (2015). Deep learn-
ing. Nature, 521(7553):436–444.
Luca, E. D. (2011). Le R
´
epertoire de la Com
´
edie-Italienne
(1716-1762).
Matthews, B. W. (1975). Comparison of the predicted and
observed secondary structure of t4 phage lysozyme.
Biochimica et Biophysica Acta (BBA)-Protein Struc-
ture, 405(2):442–451.
Moysset, B., Louradour, J., Kermorvant, C., and Wolf, C.
(2016). Learning text-line localization with shared
and local regression neural networks. In Frontiers in
Handwriting Recognition (ICFHR), 2016 15th Inter-
national Conference on, pages 1–6. IEEE.
Nair, V. and Hinton, G. E. (2010). Rectified linear units
improve restricted boltzmann machines. In Proceed-
ings of the 27th international conference on machine
learning (ICML-10), pages 807–814.
Pan, S. J. and Yang, Q. (2010). A survey on transfer learn-
ing. IEEE Transactions on knowledge and data engi-
neering, 22(10):1345–1359.
Schmidhuber, J. (2015). Deep learning in neural networks:
An overview. Neural networks, 61:85–117.
Viard-Gaudin, C., Lallican, P. M., Knerr, S., and Binter,
P. (1999). The ireste on/off (ironoff) dual handwrit-
ing database. In Document Analysis and Recognition,
1999. ICDAR’99. Proceedings of the Fifth Interna-
tional Conference on, pages 455–458. IEEE.
Transfer Learning for Structures Spotting in Unlabeled Handwritten Documents using Randomly Generated Documents
425