Learn More
We propose deeply-supervised nets (DSN), a method that simultaneously minimizes classification error and improves the directness and transparency of the hidden layer learning process. We focus our attention on three aspects of traditional convolutional-neural-network-type (CNN-type) architectures: (1) transparency in the effect intermediate layers have on(More)
We seek to improve deep neural networks by generalizing the pooling operations that play a central role in current architectures. We pursue a careful exploration of approaches to allow pooling to learn and to adapt to complex and variable patterns. The two primary directions lie in (1) learning a pooling function via (two strategies of) combining of max and(More)
One of the most promising ways of improving the performance of deep convolutional neural networks is by increasing the number of convolutional layers. However, adding layers makes training more difficult and computationally expensive. In order to train deeper networks, we propose to add auxiliary supervision branches after certain intermediate layers during(More)
We present a novel algorithmic approach to track multiple cell junctions automatically in the developing epidermis of the C. elegans embryo. 3D cell boundaries are projected into 2D for segmentation using active contours with a non-intersection force, and subsequently tracked using SIFT (Scale-Invariant Feature Transform) flow. Our method achieves MAD (Mean(More)
Effective information retrieval (IR) using domain knowledge and semantics is one of the major challenges in IR. In this paper we propose a framework that can facilitate image retrieval based on a sharable domain ontology and thesaurus. In particular, case-based learning (CBL) using a natural language phrase parser is proposed to convert a natural language(More)
We present a new feature representation method for scene text recognition problem, particularly focusing on improving scene character recognition. Many existing methods rely on Histogram of Oriented Gradient (HOG) or part-based models, which do not span the feature space well for characters in natural scene images, especially given large variation in fonts(More)
We present a framework of utilizing sharable domain ontology and thesaurus to help the retrieval of historical images of the First Emperor of China's terracotta warriors and horses. Incorporating the sharable domain ontology in RDF and RDF schemas of semantic web and a thesaurus, we implement methods to allow easily annotating images into RDF instances and(More)