Asml Netherlands B.v.

Meta Platforms Technologies, LLC

Black Sesame Technologies Inc.

Qiang Zhang

Chen Zhang

Jen-Shiang Wang

Dong Yang

Theresa Loney Casarez

Jiao Liang

Yunbo Guo

Wen Song

Yen-Wen Lu

Zheng Qi

Mu Feng

Qian Zhao

Wei Zuo

Chenhao Fang

The present invention discloses a graph partitioning system for running neural networks on resource constrained hardware systems. The graph partitioning system used for partitioning a neural network graph into a series of sub-graphs and further allow the multiple sub-graphs to be executed in available hardware subsystems. The system based on cost function as estimated computation time and memory bandwidth of partitioned sub-graphs. The graph partitioning system is a cycle estimation model of hardware that can run fast and parameterize memory latency. The graph partitioning system supports heterogeneous partition for different type accelerators such as CPU, GPU, ASIC. The present invention also discloses a method for partitioning neural network graph in to series of sub-graphs.

Neural networks graph partitioning system and method for the same

A method for evaluating images of a printed pattern. The method includes obtaining a first averaged image of the printed pattern, where the first averaged image is generated by averaging raw images of the printed pattern. The method also includes identifying one or more features of the first averaged image. The method further includes evaluating the first averaged image, using an image quality classification model and based at least on the one or more features. The evaluating includes determining, by the image quality classification model, whether the first averaged image satisfies a metric.

Utilize machine learning in selecting high quality averaged SEM images from raw images automatically

A method for improving a process model for a patterning process, the method including obtaining a) a measured contour from an image capture device, and b) a simulated contour generated from a simulation of the process model. The method also includes aligning the measured contour with the simulated contour by determining an offset between the measured contour and the simulated contour. The process model is calibrated to reduce a difference, computed based on the determined offset, between the simulated contour and the measured contour.

Method for improving a process for a patterning process

Aspects of the present disclosure are directed to a multi-sensor don/doff detection system for an artificial reality device headset. The multi-sensor don/doff detection system can use a combination of a proximity sensor, an inertial measurement unit (IMU), and an eye tracking/face tracking (ET/FT) unit to make these determinations. However, when both the ET/FT system and proximity sensor system are active, they can have system coexistence issues. Thus, only one of these systems can be used simultaneously. The multi-sensor don/doff detection system can more accurately identify don events by using input from the proximity sensor and the IMU. The multi-sensor don/doff detection system can also more accurately identify doff events by using input from the IMU and either A) the proximity sensor when the ET/FT system is disabled or B) the ET/FT system when the ET/FT system is enabled.

Artificial reality device headset DONN and DOFF detection

A method for improving a process model by measuring a feature on a printed design that was constructed based in part on a target design is disclosed. The method includes obtaining a) an image of the printed design from an image capture device and b) contours based on shapes in the image. The method also includes identifying, by a pattern recognition program, patterns on the target design that include the feature and determining coordinates, on the contours, that correspond to the feature. The method further includes improving the process model by at least a) providing a measurement of the feature based on the coordinates and b) calibrating the process model based on a comparison of the measurement with a corresponding feature in the target design.

Utilize pattern recognition to improve SEM contour measurement accuracy and stability automatically

Growing community of inventors

Campbell, CA, United States of America

Qiang Zhang