Header

Hero Section

Blog

Welcome to the NEC Labs America blog, where we cover important topics related to our research in a more informal medium than a research paper.

Project Card Grid

Check out our #computervision papers at #CVPR2022

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) is the premier annual computer vision event comprising the main conference and several co-located workshops and short courses.NEC Laboratories Media Analytics department is excited to be presenting at CVPR 2022, being held in New Orleans, LA from Sunday, June 19 to Friday, June 24, 2022.Learning To Learn Across Diverse Data Biases in Deep Face Recognition- Authors: Chang Liu, Xiang Yu, Yi-Hsuan Tsai, Ramin Moslemi, Masoud …

READ MORE

Media Analytics Research at CVPR 2020

August 9, 2020:Private-kNN: Making Differential Privacy Practical for Computer Vision Differential privacy (DP) prevents an adversary from inspecting model outputs to determine individual members of a private dataset. Consider a model A trained with the private data, while model B is trained using an additional data sample X. With differential privacy, we have a theoretical guarantee that with a certain probability, the behavior of the two models appears the same to an adversary, who may not de…

READ MORE

AI Solutions With a Guarantee of Privacy

August 17, 2020:SummaryWith increasing ethical and legal concerns about privacy for deep models in visual recognition, differential privacy has emerged as a mechanism to disguise membership of sensitive data in training datasets. Previous methods, like private aggregation of teacher ensembles (PATE), leverage a large ensemble of teacher models trained on disjoint subsets of private data to transfer knowledge to a student model with privacy guarantees. However, labeled data is often expensive, a…

READ MORE

Understanding Road Scenes in Videos

August 24, 2020:Buyu LiuBingbing ZhuangSamuel SchulterManmohan ChandrakerSummaryGiven a video of a complex scene, we represent various elements of the scene in a way that is both interpretable and accurate. This allows explainable decision-making for downstream application and intuitive visualization for human-machine interaction. For a road scene, we predict elements of the layout such as the number of lanes, positions of crosswalks, distance from the intersection and so on. These predictions …

READ MORE

Privacy-Aware AI Models

January 6, 2021:Differential PrivacyOur differential privacy method provides customers a provable guarantee of privacy against all types of attacks and for all types of data. Prior methods from Google achieve privacy by splitting the private data and training an ensemble of teacher models to guide the deployed model, which needs a large amount of data. Our method relies on privacy amplification by sub-sampling to achieve higher accuracy and privacy, with orders of magnitude less data. On the sa…

READ MORE

Universal Representations for Responsible and Robust AI

January 6, 2021:URLearning for Semantic SegmentationAnnotation Effort vs Performance:Full Supervision vs our Approach:URLearning for Facial RecognitionOur URFace vs CosFace:CFP (Pose Variation) TinyFace (Low Resolution) IJB-A (Surveillance Situation) ConclusionFor more technical details, see our other material on universal representations for responsible and robust AI:ECCV 2020 Paper on Weak Labels for Segmentation ECCV2020 Paper on Object Detection CVPR2020 Paper on Face Recognition ICCV2019 P…

READ MORE

Camera Motion Estimation by Learning Uncertainty-Aware Fusion

June 21, 2020:"Fusing the Old with the New: Learning Relative Camera Pose with Geometry-Guided Uncertainty"Bingbing ZhuangManmohan ChandrakerSummaryHumans and autonomous agents always move in a 3D environment rather than on a 2D image plane, making 3D motion perception a fundamental capability for any intelligent system. As such, estimating relative camera pose or motion from just images has been a long-standing problem in the field of computer vision since the 1980s, and it serves as an essent…

READ MORE

Divide-and-Conquer for Lane-Aware Diverse Trajectory Prediction

June 21, 2020:"Divide and Conquer for Lane Aware Diverse Trajectory Prediction"SummaryTrajectory prediction is a safety-critical tool for autonomous vehicles to plan and execute actions. Our work addresses two key challenges in trajectory prediction: (i) learning multimodal outputs and (ii) improving predictions by imposing constraints using driving knowledge. Recent methods have achieved strong performances using multi-choice learning objectives like winner-takes-all (WTA) or best-of-many. But…

READ MORE

Cross-Domain Similarity Learning for Face Recognition in Unseen Domains

June 21, 2020: Masoud FarakiXiang YuYumin SuhManmohan ChandrakerSummaryFacial recognition (FR)-based systems continue to be the preferred approach for many applications such as biometric technologies and emotion analysis. That is mainly because FR is the most natural way to meet these types of demands with no physical interaction with the end-user. Furthermore, it is fast and easy to deploy and implement, while showing promising outcomes using deep neural networks. However, many FR methods base…

READ MORE