This work aims to compare the learning features with Convolutional Neural Networks (CNN) and the handcrafted features. In order to determine which the best between these two type of features. We consider our previous baseline HMM system [1] for Arabic handwritten word recognition. Experiments have been conducted on the well-known IFN/ENIT database. Achieved results using CNN features are better than those obtained by the hand-crafted features. This demonstrates the high efficiency of CNN results from the strong capability for hierarchical feature learning given a large amount of data. However, Hand-engineered features are not generated from an optimization process to be compatible with the specific problem, and insufficient to be encoded with supervision.
CITATION STYLE
Amrouch, M., & Rabi, M. (2018). Deep neural networks features for Arabic handwriting recognition. In Lecture Notes in Networks and Systems (Vol. 25, pp. 138–149). Springer. https://doi.org/10.1007/978-3-319-69137-4_14
Mendeley helps you to discover research relevant for your work.