arXiv subjects

Peter Zhang

Publications and source records attributed to Peter Zhang.

22 records · Page 2Linked to original sources

arXiv2021

DEFT: Detection Embeddings for Tracking

Most modern multiple object tracking (MOT) systems follow the tracking-by-detection paradigm, consisting of a detector followed by a method for associating detections into tracks. There is a long history in tracking of combining motion and appearance features to provide robustness to occlusions and other challenges, but typically this comes with the trade-off of a more complex and slower implementation. Recent successes on popular 2D tracking benchmarks indicate that top-scores can be achieved using a state-of-the-art detector and relatively simple associations relying on single-frame spatial offsets -- notably outperforming contemporary methods that leverage learned appearance features to help re-identify lost tracks. In this paper, we propose an efficient joint detection and tracking model named DEFT, or "Detection Embeddings for Tracking." Our approach relies on an appearance-based object matching network jointly-learned with an underlying object detection network. An LSTM is also added to capture motion constraints. DEFT has comparable accuracy and speed to the top methods on 2D online tracking leaderboards while having significant advantages in robustness when applied to more challenging tracking data. DEFT raises the bar on the nuScenes monocular 3D tracking challenge, more than doubling the performance of the previous top method. Code is publicly available.

cs.CV↗

arXiv2021

Model Mis-specification and Algorithmic Bias

Machine learning algorithms are increasingly used to inform critical decisions. There is a growing concern about bias, that algorithms may produce uneven outcomes for individuals in different demographic groups. In this work, we measure bias as the difference between mean prediction errors across groups. We show that even with unbiased input data, when a model is mis-specified: (1) population-level mean prediction error can still be negligible, but group-level mean prediction errors can be large; (2) such errors are not equal across groups; and (3) the difference between errors, i.e., bias, can take the worst-case realization. That is, when there are two groups of the same size, mean prediction errors for these two groups have the same magnitude but opposite signs. In closed form, we show such errors and bias are functions of the first and second moments of the joint distribution of features (for linear and probit regressions). We also conduct numerical experiments to show similar results in more general settings. Our work provides a first step for decoupling the impact of different causes of bias.

cs.LG↗

arXiv2018

Theoretical model of a finite force at the moving contact line

In theoretical analyses of the moving contact line, an infinite force along the solid wall has been reported based off the non-integrable stress along a single interface. In this investigation we demonstrate that the stress singularity is integrable and results in a finite force at the moving contact line if the contact line is treated as a one-dimensional manifold and all three interfaces that make up the moving contact line are taken into consideration. This is due to the dipole nature of the vorticity and pressure distribution around the moving contact line. Mathematically, this finite force is determined by summing all the forces that act over an infinitesimally small cylindrical control volume that encloses the entire moving contact line. With this finite force, we propose a new dynamic Young's equation for microscopic dynamic contact angle that is a function of known parameters only, specifically the interface velocity, surface tension, and fluid viscosity. We combine our model with Cox's model for apparent dynamic contact angle and find good agreement with published dynamic contact angle measurements.

physics.flu-dyn↗

arXiv2005

A Data Analysis Technique for the LIGO-ALLEGRO Stochastic Background Search

We describe the cross-correlation measurements being carried out on data from the LIGO Livingston Observatory and the ALLEGRO resonant bar detector. The LIGO data are sampled at 16384 Hz while the ALLEGRO data are base-banded, i.e., heterodyned at 899 Hz and then sampled at 250 Hz. We handle these different sampling parameters by working in the Fourier domain, and demonstrate the approximate equivalence of this measurement to a hypothetical time-domain method in which both data streams are upsampled.

gr-qc↗