Nec Laboratories America, Inc.

University of Southern California

Nec Soft, Ltd.

Fengjun Lv

Kai Hung Yu

Jinjun Wang

Yihong Gong

Shenghuo Zhu

Ming M Yang

Kazuya Ueki

Wei Xu

Ramakant Nevatia

Xi Zhou

Systems and methods are disclosed for determining personal characteristics from images by generating a baseline gender model and an age estimation model using one or more convolutional neural networks (CNNs); capturing correspondences of faces by face tracking, and applying incremental learning to the CNNs and enforcing correspondence constraint such that CNN outputs are consistent and stable for one person.

Systems and methods for determining personal characteristics

An automated human action recognition system may automatically recognize one or more actions of a human from 2D input image data representing a sequential series of input images of the human performing the one or more actions. Each input image may be from an unknown viewpoint. A computer memory system may contain 2D reference image data representing a plurality of reference actions which a human may perform. The 2D reference image data may include a plurality of linked sequences of key poses, including a linked sequence of key poses for each reference action. For each reference action, each key pose within the linked sequence of key poses for the reference action may consist essentially of 2D image data that is representative of a human figure performing the reference action at a selected point during the reference action. The timing of the selected points within the linked sequence of key poses for the reference action may be based on changes in the position of the human figure during the performance of the reference action. The linked sequence of key poses for the reference action may uniquely distinguish it from the linked sequence of key poses for all of the other reference actions. A computer processing system may be configured to determine which of the reference actions best matches the 2D input image data with no knowledge of the viewpoint of the 2D input image data.

Automated single viewpoint human action recognition by matching linked sequences of key poses

Systems and methods are disclosed to perform image parsing on one or more images by identifying a set of similar regions from each image; assigning one or more region labels to each region and generating multiple hypotheses for region label assignment; and detecting class, location and boundary of each object in the image, wherein object classification, detection and segmentation are performed jointly during image parsing.

Fast image parsing by graph adaptive dynamic programming (GADP) performing classification, detection, and segmentation simultaneously

A human tracking apparatus and method capable of highly accurately tracking the movement of persons photographed in moving images includes: an image memorythat stores an inputted frame image; a human detecting unitthat detects persons photographed in the inputted frame image; a candidate registering unitthat registers already detected persons as candidates; a similarity index calculating unitthat calculates similarity indices indicating the similarity between the persons detected in the inputted frame image and the registered candidates for two or more types of parameters based on the stored frame images in relation to all combinations of the persons and the candidates; a normalizing unitthat normalizes the similarity indices; an integrating unitthat integrates the normalized indices for each combination of the detected persons and the candidates; and a tracking unitthat identifies a person the same as an arbitrary candidate based on the similarity indices.

Human tracking apparatus, human tracking method, and human tracking processing program

Systems and methods are disclosed for classifying an input image by detecting one or more feature points on the input image; extracting one or more descriptors from each feature point; applying a codebook to quantize each descriptor and generate code from each descriptor; applying spatial pyramid matching to generate histograms; and concatenating histograms from all sub-regions to generate a final representation of the image for classification.

Locality-constrained linear coding systems and methods for image classification

System and methods are disclosed to perform multi-human 3D tracking with a plurality of cameras. At each view, a module receives each camera output and provides 2D human detection candidates. A plurality of 2D tracking modules are connected to the CNNs, each 2D tracking module managing 2D tracking independently. A 3D tracking module is connected to the 2D tracking modules to receive promising 2D tracking hypotheses. The 3D tracking module selects trajectories from the 2D tracking modules to generate 3D tracking hypotheses.

Growing community of inventors

San Jose, CA, United States of America

Fengjun Lv