TY - GEN
T1 - A further step to perfect accuracy by training CNN with larger data
AU - Uchida, Seiichi
AU - Ide, Shota
AU - Iwana, Brian Kenji
AU - Zhu, Anna
PY - 2016/7/2
Y1 - 2016/7/2
N2 - Convolutional Neural Networks (CNN) are on the forefront of accurate character recognition. This paper explores CNNs at their maximum capacity by implementing the use of large datasets. We show a near-perfect performance by using a dataset of about 820,000 real samples of isolated handwritten digits, much larger than the conventional MNIST database. In addition, we report a near-perfect performance on the recognition of machine-printed digits and multi-font digital born digits. Also, in order to progress toward a universal OCR, we propose methods of combining the datasets into one classifier. This paper reveals the effects of combining the datasets prior to training and the effects of transfer learning during training. The results of the proposed methods also show an almost perfect accuracy suggesting the ability of the network to generalize all forms of text.
AB - Convolutional Neural Networks (CNN) are on the forefront of accurate character recognition. This paper explores CNNs at their maximum capacity by implementing the use of large datasets. We show a near-perfect performance by using a dataset of about 820,000 real samples of isolated handwritten digits, much larger than the conventional MNIST database. In addition, we report a near-perfect performance on the recognition of machine-printed digits and multi-font digital born digits. Also, in order to progress toward a universal OCR, we propose methods of combining the datasets into one classifier. This paper reveals the effects of combining the datasets prior to training and the effects of transfer learning during training. The results of the proposed methods also show an almost perfect accuracy suggesting the ability of the network to generalize all forms of text.
UR - http://www.scopus.com/inward/record.url?scp=85012874404&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85012874404&partnerID=8YFLogxK
U2 - 10.1109/ICFHR.2016.0082
DO - 10.1109/ICFHR.2016.0082
M3 - Conference contribution
T3 - Proceedings of International Conference on Frontiers in Handwriting Recognition, ICFHR
SP - 405
EP - 410
BT - Proceedings - 2016 15th International Conference on Frontiers in Handwriting Recognition, ICFHR 2016
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 15th International Conference on Frontiers in Handwriting Recognition, ICFHR 2016
Y2 - 23 October 2016 through 26 October 2016
ER -