Spatial Transformer Networks for Curriculum Learning

TL;DR

This paper proposes using Spatial Transformer Networks to generate easier training tasks in curriculum learning, demonstrating improved accuracy on cluttered MNIST datasets by leveraging image preprocessing to facilitate neural network training.

Contribution

The work introduces a novel approach to curriculum learning by utilizing Spatial Transformer Networks to create easier tasks, enhancing training efficiency and accuracy.

Findings

Achieved 3.8 percentage points improvement in accuracy on cluttered MNIST

Validated the effectiveness of STN-based curriculum on image classification tasks

Showed that processed images serve as effective easier tasks for training

Abstract

Curriculum learning is a bio-inspired training technique that is widely adopted to machine learning for improved optimization and better training of neural networks regarding the convergence rate or obtained accuracy. The main concept in curriculum learning is to start the training with simpler tasks and gradually increase the level of difficulty. Therefore, a natural question is how to determine or generate these simpler tasks. In this work, we take inspiration from Spatial Transformer Networks (STNs) in order to form an easy-to-hard curriculum. As STNs have been proven to be capable of removing the clutter from the input images and obtaining higher accuracy in image classification tasks, we hypothesize that images processed by STNs can be seen as easier tasks and utilized in the interest of curriculum learning. To this end, we study multiple strategies developed for shaping the training curriculum, using the data generated by STNs. We perform various experiments on cluttered MNIST and Fashion-MNIST datasets, where on the former, we obtain an improvement of $3.8$pp in classification accuracy compared to the baseline.