Intel Corporation

Pan Asia Technical Automotive Center Company Limited

Saic General Motors Corporation Limited

Yan Hao

Kun Yu

Lu Li

Zhi Yong Zhu

Feng Chen

Xiaoming Chen

Ciyong Chen

Hui Li

Yang Liu

Tao Li

Jin Fu

Qiang Zhou

Zhiyong Zhu

Xiaotian Guo

Jianguo Pei

Junqiao Peng

Mu Qian

Zhankui He

An example apparatus for detecting objects in video frames includes a receiver to receive a plurality of video frames from a video camera. The apparatus also includes a first still image object detector to receive a first frame of the plurality of video frames and calculate localization information and confidence information for each potential object patch in the first frame. The apparatus further includes a second still image object detector to receive an adjacent frame of the plurality of video frames adjacent to the first frame and calculate localization information and confidence information for each potential object patch in the adjacent frame. The apparatus includes a similarity detector trained to detect paired patches between the first frame and the adjacent frame based on a comparison of the detected potential object patches. The apparatus further includes an enhancer to modify a prediction result for a paired patch in the adjacent frame to a prediction result of a corresponding paired patch in the first frame including a higher confidence score than the prediction result of the paired patch in the adjacent frame.

Detecting objects in video frames using similarity detectors

Techniques related to training and implementing a bidirectional pairing architecture for object detection are discussed. Such techniques include generating a first enhanced feature map for each frame of a video sequence by processing the frames in a first direction, generating a second enhanced feature map for frame by processing the frames in a second direction opposite the first, and determining object detection information for each frame using the first and second enhanced feature map for the frame.

Bidirectional pairing architecture for object detection in video

System and techniques are provided for three-dimension (3D) semantic segmentation. A device for 3D semantic segmentation includes: an interface, to obtain a point cloud data set for a time-ordered sequence of 3D frames, the 3D frames including a current 3D frame and one or more historical 3D frames previous to the current 3D frame; and processing circuitry, to: invoke a first artificial neural network (ANN) to estimate a 3D scene flow field for each of the one or more historical 3D frames by taking the current 3D frame as a reference frame; and invoke a second ANN to: produce an aggregated feature map, based on the reference frame and the estimated 3D scene flow field for each of the one or more historical 3D frames; and perform the 3D semantic segmentation based on the aggregated feature map.

Point cloud based 3D semantic segmentation

The present invention discloses a method and system for unlocking vehicle engine cover. Method comprising: acquiring an action signal transmitted from an action sensor, and executing an action signal judging step; executing an unlocking step if the received action signal matches a preset unlock action signal, otherwise not executing the unlocking step, the unlock action signal is a combination of a signal or a plurality of consecutive signals; unlocking step: sending an unlock signal to an engine cover lock, wherein the unlock signal is used to control the engine cover lock to perform an unlock operation on a vehicle engine cover. The present invention detects an action signal from an action sensor, and controls whether or not to unlock the vehicle engine cover by matching the action signal with a preset unlock action signal. So users do not need to enter into the passenger compartment for unlocking. The operation is both convenient and rich sense of technology.

Method and system for unlocking vehicle engine cover

In accordance with some embodiments, spatial and temporal locality between threads executing on graphics processing units may be analyzed and tracked in order to improve performance. In some applications where a large number of threads are executed and those threads use common resources such as common data, affinity tracking may be used to improve performance by reducing the cache miss rate and to more effectively use relatively small-sized caches.

Scheduling thread execution based on thread affinity

Growing community of inventors

Shanghai, China

Yan Hao