Graphic notation is an essential means of communication.
Such notation includes diagrams, mathematical expression, and chemical equations.
In this project, we investigate machine learning techniques for
recognition of handwritten graphics and graph-based methods for automatic annotation
of ground-truth in graphics.
A handwritten mathematical expression (left) and its structural components (right)
Image operator learning with Deep Learning
In this project, we investigate Deel Learning techniques for
automatic learning of image operators. We model the image operator learning
task as a problem of learning Convolutional Neural Network (CNN) classifiers.
The learning framework considers operators that are translation-invariant
and locally defined with respect to a finite non-empty window. The operators
can be applied to a variety of tasks, including text segmentation, staff line removal
in music score images, and object detection.