Black Sesame Technologies Inc.

Utah State University

Black Sesame International Holding Limited

Sesame Technologies Inc.

Qun Gu

Zuoguan Wang

Lei Zhang

Scott Ezra Budge

Jizhang Shan

Zheng Qi

Xiaomin Liu

Tian Zhou

Yilin Song

Zilong Hu

Zheng Li

Chengyu Xiong

Meng Meng

Liying Liu

Chenghao Zhang

A method of human pose estimation, including, receiving an image frame, extracting in at least a first stage a first stage image feature based on the image frame, extracting in at least a second stage a second stage image feature based on the first stage image feature, extracting in at least a subsequent stage a subsequent stage image feature based on the second stage image feature, up-sampling the subsequent stage image feature, up-sampling the second stage image feature, concatenating the first stage image feature, the up-sampled second stage image feature and the up-sampled subsequent stage image feature and outputting a feature map based on an output of the concatenation.

Method of human pose estimation

An example fused convolutional layer, comprising, a comparator capable of reception of a first zero point and a multiply-accumulation result, a first multiplexer coupled to the comparator, wherein the first multiplexer receives a plurality of power-of-two exponent values, a shift normalizer, coupled to the first multiplexer, wherein the shift normalizer is capable of receiving the multiply-accumulation result and the plurality of power-of-two exponent values, wherein the shift normalizer limits a quantization of the multiply-accumulation result to a power-of-two scale and a second multiplexer coupled to an output of the shift normalizer, the first multiplexer and receives a second zero point and outputs an activation.

Integer-based fused convolutional layer in a convolutional neural network

A method of face anti-spoofing, comprising, receiving a near infra-red facial image, having a near infrared channel, receiving a red-green-blue facial image, having a red channel, a green channel and a blue channel, generating a synthetic three channel image based on the near infrared channel, the red channel, the green channel and the blue channel and training a deep neural network based on the synthetic three channel image.

RGB-NIR dual camera face anti-spoofing method

A method of age and gender estimation, comprising receiving an input image, detecting a facial image within the input image, estimating a head pose based on a set of facial image intensities of the facial image, wherein the head pose is expressed as a yaw, a pitch and a roll, determining whether the yaw, the pitch and the roll of the head pose is less than a predetermined threshold, aligning the facial image if the yaw, the pitch and the roll of the head pose are less than the predetermined threshold and predicting an age and a gender of the aligned facial image.

Age and gender estimation using a convolutional neural network

A method of hybrid lane modeling, including, receiving a roadway image, extracting a set of lane points from the roadway image, fitting a polynomial line to the set of lane points, determining a fitted error of the fitted polynomial line, outputting the polynomial line if the fitted error is less than a predetermined threshold, selecting a set of clean lane points from the set of lane points if the fitted error is greater than the predetermined threshold and interpolating a cubic spline line to the set of clean lane points.

Hybrid lane model

A method of describing a temporal event, including receiving a video sequence of the temporal event, extracting at least one physical characteristic of an at least one occupant within the video sequence, extracting at least one action of the at least one occupant within the video sequence, extracting at least one interaction of the at least one occupant with a secondary occupant within the video sequence, determining a safety level of the temporal event within a vehicle based on at least one of the at least one action and the at least one interaction and describing the at least one physical characteristic of the at least one occupant and at least one of the at least one action and the at least one interaction of the at least one occupant.

In-cabin occupant behavoir description

A method of pruning a convolutional neural network, comprising at least one of determining a number of channels (N) between a network input and a network output, constructing N lookup tables, each lookup table matched to a respective channel and pruning filters in the convolutional neural network to create a shortcut between the network input and the network output based on the N lookup tables.

Method of structured network pruning and sparsity speed-up

A system on a chip, including a first domain having a first processor, a first local memory coupled to the first processor, wherein the first local memory having a first memory format and a first sub-network coupled to the first processor, a second domain having a second processor, a second local memory coupled to the second processor and a second sub-network coupled to the second processor, wherein the second local memory having a second memory format which differs from the first memory format, a multi-tier network coupled to the first sub-network and the second sub-network, a global memory coupled to the multi-tier network and a multi-port DDR controller coupled to the global memory to receive, transmit and share the first local memory having the first memory format and the second local memory having the second memory format based on a predetermined criteria.

Heterogeneous computation and hierarchical memory image sensing pipeline

A method of adaptive quantization for a convolutional neural network, includes at least one of receiving an acceptable model accuracy, determining a float value multiply accumulate for the layer based on a float value weight and a float value input, quantizing the float value weight at multiple weight quantization precisions, quantizing the float value input at multiple input quantization precisions, determining a multiply accumulate at multiple multiply accumulate quantization precisions based on the weight quantization precisions and the input quantization precisions, determining multiple quantization errors based on differences between the float value multiply accumulate and the multiple multiply accumulate quantization precisions and selecting one of the multiple weight quantization precisions, one of the multiple input quantization precisions and one of the multiple multiply accumulate quantization precisions based on the predetermined acceptable model accuracy and the multiple quantization errors.

Adaptive quantization and mixed precision in a network

A method of distant on-road object detection, comprising, capturing an initial frame of a roadway lane in a direction of travel, detecting lane edges in the initial frame, capturing a subsequent frame of the roadway lane, cropping the subsequent frame based on the detected lane edges of the initial frame resulting in an image patch, detecting an image patch object within the image patch, down-sampling the subsequent frame, detecting a down-sampled object within the down-sampled subsequent frame and merging the detections within the image patch and the detections within the down-sampled subsequent frame.

Distant on-road object detection

A method of parking lot tracking including receiving a plurality of camera images from a plurality of cameras attached to a vehicle in motion, stitching the plurality of camera images to simulate a surround view of the vehicle, recognizing at least one potential parking space within the surround view of the vehicle, estimating a motion parameter by camera motion estimation of the plurality of cameras and tracking the at least one potential parking space based on the motion parameter.

Real-time robust surround view parking space detection and tracking

A method of ultra-low rate video encoding, including capturing a frame within a video stream, capturing a global position of the frame, capturing a heading of the frame, linking the global position and the heading to the frame, determining a foreground of the frame based on the global position of the frame and the heading of the frame, determining an at least one region of interest within the foreground, analyzing the at least one region of interest, removing a background from the frame based on the global position and the heading, wherein the background is complementary to the foreground and encoding the foreground, the global position, the heading and the at least one region of interest.

Ultra-low bitrate coding based on 3D map reconstruction and decimated sub-pictures

Methods and systems for encoding, transmitting and decoding digital images by rate-distortion adaptive zerotree-based residual vector quantization are disclosed. A method of the invention includes receiving a digital image, transforming the digital image into the wavelet domain generating a pyramid hierarchy, losslessly encoding a top LL subband from the pyramid hierarchy, encoding other subbands by vector quantization based on a zerotree insignificance prediction, generating an encoded image from the lossless encoding and vector quantization encoding, transmitting the encoded image along a communications channel, receiving the encoded image transmitted along the communications channel, reconstructing a zerotree from the encoded image, vector quantization decoding subbands from the encoded image other than a top LL subband, losslessly decoding the top LL subband from the encoded image, reverse wavelet transforming the top LL subband and the vector quantization decoded subbands and outputting a decoded image.

Growing community of inventors

San Jose, CA, United States of America

Qun Gu