TextCaps : Handwritten Character Recognition with Very Small Datasets
This addresses the challenge of applying deep learning to localized languages that lack substantial labeled training data, though it is incremental as it builds on existing augmentation and loss function techniques.
The paper tackles the problem of handwritten character recognition for languages with limited labeled data by generating realistic augmented samples from very small datasets, achieving state-of-the-art results with only 200 training samples per class on the EMNIST-letter dataset and matching existing results on other datasets.
Many localized languages struggle to reap the benefits of recent advancements in character recognition systems due to the lack of substantial amount of labeled training data. This is due to the difficulty in generating large amounts of labeled data for such languages and inability of deep learning techniques to properly learn from small number of training samples. We solve this problem by introducing a technique of generating new training samples from the existing samples, with realistic augmentations which reflect actual variations that are present in human hand writing, by adding random controlled noise to their corresponding instantiation parameters. Our results with a mere 200 training samples per class surpass existing character recognition results in the EMNIST-letter dataset while achieving the existing results in the three datasets: EMNIST-balanced, EMNIST-digits, and MNIST. We also develop a strategy to effectively use a combination of loss functions to improve reconstructions. Our system is useful in character recognition for localized languages that lack much labeled training data and even in other related more general contexts such as object recognition.