Advisor: Dr. Kosta Derpanis


Master's thesis This thesis introduces a method to both obtain segmentation information and integrate it uniformly within a convolutional neural network (CNN). The resulting network is called a segmentation-aware CNN, because the network can change its behaviour at each image location according to local segmentation cues.
ICLR 2015 This paper proposes a new deep convolutional neural network architecture that learns pixel embeddings, such that pairwise distances between the embeddings can be used to infer semantic similarity of the underlying regions.
ICDAR 2015 This paper presents a new state-of-the-art for document image classification and retrieval, using features learned by deep convolutional neural networks (CNNs).
Best Student Paper Award
RVL-CDIP The RVL-CDIP (Ryerson Vision Lab Complex Document Information Processing) dataset consists of 400,000 grayscale images in 16 classes, with 25,000 images per class. There are 320,000 training images and 80,000 test images. The images are sized so their largest dimension does not exceed 1000 pixels.
Neural network visualization This paper presents a new interactive visualization of neural networks trained on handwritten digit recognition, with the intent of showing the actual behavior of the network given user-provided input. The user can interact with the network through a drawing pad, and watch the activation patterns of the network respond in real time.
Featured in Popular Science
Demystifying This is a tutorial-style document that explores the mathematics of deep convolutional neural networks. It examines parameter tuning, fully-connected networks, error minimization, back-propagation, convolutional networks, and finally deep networks. (This is a work in progress.)

Research Interests

I am currently working on image understanding tasks, using neural networks and deep learning. I am also interested in data hallucination, 3D vision, and video understanding.