S
Siqi Deng
Researcher at Amazon.com
Publications - 12
Citations - 53
Siqi Deng is an academic researcher from Amazon.com. The author has contributed to research in topics: Computer science & Facial recognition system. The author has an hindex of 2, co-authored 5 publications receiving 12 citations.
Papers
More filters
Posted Content
Positive-Congruent Training: Towards Regression-Free Model Updates
TL;DR: A simple approach for PC training, Focal Distillation, is proposed, which enforces congruence with the reference model by giving more weights to samples that were correctly classified, and it is found that negative flips can be further reduced without affecting the new model’s accuracy.
Proceedings ArticleDOI
Positive-Congruent Training: Towards Regression-Free Model Updates
TL;DR: In this article, the authors propose a positive-congruent (PC) training method to reduce negative flips by giving more weights to samples that were correctly classified by the reference model.
Posted Content
SMOT: Single-Shot Multi Object Tracking.
TL;DR: A new tracking framework that converts any single-shot detector model into an online multiple object tracker, which emphasizes simultaneously detecting and tracking of the object paths, which achieves state-of-the-art performance.
Proceedings ArticleDOI
The Caltech Fish Counting Dataset: A Benchmark for Multiple-Object Tracking and Counting
Justin Kay,Peter Kulits,Suzanne C. Stathatos,Siqi Deng,Erik Young,Sara Beery,Grant Van Horn,Pietro Perona +7 more
TL;DR: The Caltech Fish Counting Dataset (CFC), a large-scale dataset for detecting, tracking, and counting fish in sonar videos, is presented, which allows researchers to train MOT and counting algorithms and evaluate generalization performance at unseen test locations.
Book ChapterDOI
Unsupervised and Semi-supervised Bias Benchmarking in Face Recognition
TL;DR: In this article , a semi-supervised performance evaluation for face recognition (SPE-FR) method is proposed, which is based on parametric Bayesian modeling of the face embedding similarity scores.