Previous part-based attribute recognition approaches perform part detection and attribute recognition in separate steps. The parts are not optimized for attribute recognition and therefore could be sub-optimal. We present an end-to-end deep learning approach to overcome the limitation. It generates object parts from key points and perform attribute recognition accordingly, allowing adaptive spa...

Using a recent improvement by Bettin and Chandee to a bound of Duke, Friedlander and Iwaniec (1997) on double exponential sums with Kloosterman fractions, we establish a uniformity of distribution result for the fractional parts of Dedekind sums s(m,n) with m and n running over rather general sets. Our result extends earlier work of Myerson (1988) and Vardi (1987). Using different techniques, w...

In its original form [D], Dyson’s Lemma bounds the order of vanishing of a polynomial in two variables at a finite set of points and was used in studying rational approximations to a fixed algebraic irrational number. Bombieri [B] revived interest in Dyson’s Lemma and subsequently Esnault and Viehweg [EV] and Vojta [V1] vastly extended the scope of the result, to polynomials in several variable...

We develop a new high order accurate time-integration technique for initial value problems. We focus on problems that originate from a space approximation using high order finite difference methods on summation-by-parts form with weak boundary conditions, and extend that technique to the timedomain. The new time-integration method is global, high order accurate, unconditionally stable and toget...

This work presents a parts-based person detection framework. The individual body segmentslegs, torsos, and heads are detected by Adaboost classifiers, which have proven useful other classification tasks [12] [6] [1] [13]. The body segment candidates are combined into kinematically plausible configurations and each is assigned a score. The body configurations are calculated using an efficient dy...

This paper describes a system (ParTS) for automatic hardware/software partitioning of applications described in the concurrent programming language occam. Based on algebraic transformations of occam programs, the strategy guarantees, by construction, that the partitioning process preserves the semantics of the original description. ParTS has been developed as an extension of OTS — a tool implem...

A shape representation is presented that has been shown competent to accurately describe an extensive variety of natural forms (e g., people, mountains, clouds, trees), as well as man-made forms, in a succinct and natural manner. The ap preach taken in this representational system is to describe scene structure at a scale that is similar to our naive perceptual notion of “a part,’ by use of des...

Programmable parts orienting is an important capability for exible automation systems. Here we study how a part grasped in an unknown orientation by a force-controlled robot can be oriented by a sequence of push-align actions against a wall followed by sensor measurements of the distance from the grasp point to the wall. This paper concentrates on three issues: planning a sequence of actions to...

In this paper, we propose a new framework for action localization that tracks people in videos and extracts full-body human tubes, i.e., spatio-temporal regions localizing actions, even in the case of occlusions or truncations. This is achieved by training a novel human part detector that scores visible parts while regressing full-body bounding boxes. The core of our method is a convolutional n...