-
1.发明申请公开(公告)号：US20200302276A1
公开(公告)日：2020-09-24
申请号：US16586500
申请日：2019-09-27
发明人： Lin Yang , Bin Yang , Hua Zhou , Xiaochun Li , Wenhan Zhang , Qi Dong , Yequn Zhang , Yongxiong Ren , Patrick Dong
摘要： An artificial intelligence (AI) semiconductor having an embedded convolution neural network (CNN) may include a first convolution layer and a second convolution layer, in which the weights of the first layer and the weights of the second layer are quantized in different bit-widths, thus at different compression ratios. In a VGG neural network, the weights of a first group of convolution layers may have a different compression ratio than the weights of a second group of convolution layers. The weights of the CNN may be obtained in a training system including convolution quantization and/or activation quantization. Depending on the compression ratio, the weights of a convolution layer may be trained with or without re-training. An AI task, such as image retrieval, may be implemented in the AI semiconductor having the CNN described above.
-
公开(公告)号：US10296817B1
公开(公告)日：2019-05-21
申请号：US15941514
申请日：2018-03-30
发明人： Lin Yang , Patrick Z. Dong , Charles Jin Young , Jason Dong , Wenhan Zhang , Baohua Sun
摘要： Apparatus for recognition of handwritten Chinese characters contains a bus, an input means connecting to the bus for receiving input imagery data created from a handwritten Chinese character, a Cellular Neural Networks or Cellular Nonlinear Networks (CNN) based integrated circuit operatively connecting to the bus for extracting features out of the input imagery data using pre-trained filter coefficients of a plurality of order convolutional layers stored therein, a memory connecting the bus, the memory being configured for storing weight coefficients of fully-connected (FC) layers, a processing unit connecting to the bus for performing computations of FC layers to classify the extracted features from the CNN based integrated circuit to a particular Chinese character in a predefined Chinese character set, and a display unit connecting to the bus for displaying the particular Chinese character. Greater than 95% recognition accuracy is achieved using multiple bi-valued 3×3 filter kernels as pre-trained filter coefficients.
-
公开(公告)号：US20210019602A1
公开(公告)日：2021-01-21
申请号：US16516229
申请日：2019-07-18
发明人： Lin Yang , Baohua Sun , Yongxiong Ren , Wenhan Zhang
摘要： An integrated circuit may include multiple cellular neural networks (CNN) processing engines coupled in a loop circuit and configured to perform an AI task. Each CNN processing engine includes multiple convolution layers, a first memory buffer to store imagery data and a second memory buffer to store filter coefficients. The CNN processing engines are configured to perform convolution operations over an input image simultaneously in one or more iterations. In each iteration, various sub-images of the input image are loaded to the first memory buffer circularly. A portion of the filter coefficients corresponding to the sub-image are loaded to the second memory buffer in a cyclic order. Data may be arranged in the second memory buffer to facilitate loading of duplicate filter coefficients among at least two convolution layers without requiring duplicate memory space. Methods of training a CNN model having duplicate weights are also provided.
-
公开(公告)号：US10733039B2
公开(公告)日：2020-08-04
申请号：US16230020
申请日：2018-12-21
发明人： Chyu-Jiuh Torng , Daniel H. Liu , Wenhan Zhang , Hualiang Yu
摘要： This disclosure relates to testing of integrated artificial intelligence (AI) circuit with embedded memory to improve effective chip yield and to mapping addressable memory segments of the embedded memory to multilayer AI networks at the network level, layer level, parameter level, and bit level based on bit error rate (BER) of the addressable memory segments. The disclosed methods and systems allows for deployment of one or more multilayer AI networks in an AI circuit with sufficient model accuracy even when the embedded memory has an overall BER higher than a preferred overall threshold.
-
公开(公告)号：US10387740B2
公开(公告)日：2019-08-20
申请号：US15984334
申请日：2018-05-19
发明人： Lin Yang , Patrick Z. Dong , Charles Jin Young , Jason Z. Dong , Wenhan Zhang , Baohua Sun
摘要： A deep learning object detection and recognition system contains a number of cellular neural networks (CNN) based integrated circuits (ICs) operatively coupling together via the network bus. The system is configured for detecting and then recognizing one or more objects out of a two-dimensional (2-D) imagery data. The 2-D imagery data is divided into N set of distinct sub-regions in accordance with respective N partition schemes. CNN based ICs are dynamically allocated for extracting features out of each sub-region for detecting and then recognizing an object potentially contained therein. Any two of the N sets of sub-regions overlap each other. N is a positive integer. Object detection is achieved with a two-category classification using a deep learning model based on approximated fully-connected layers, while object recognition is performed using a local database storing feature vectors of known objects.
-
6.发明申请公开(公告)号：US20190220699A1
公开(公告)日：2019-07-18
申请号：US15871941
申请日：2018-01-15
发明人： Xiang Gao , Lin Yang , Wenhan Zhang
IPC分类号： G06K9/62 , H04N19/186 , G06N3/04 , G06N3/08
CPC分类号： G06K9/6256 , G06N3/04 , G06N3/08 , H04N19/186
摘要： Methods of encoding image data for loading into an artificial intelligence (AI) integrated circuit are provided. The AI integrated circuit may have an embedded cellular neural network for implementing AI tasks based on the loaded image data. An encoding method may include: using in input image to generate a plurality of output images, wherein each pixel in the input image is approximated by a combination of values of corresponding pixels in the output images; loading the plurality of output images into the AI chip; executing programming instructions contained in the AI chip to generate an image recognition result based on the at least one of the plurality of output images; and outputting the image recognition result. The encoding method also trains a convolution neural network (CNN) and loads the weights of the CNN into the AI integrated circuit for implementing the AI tasks.
-
公开(公告)号：US20180247113A1
公开(公告)日：2018-08-30
申请号：US15963990
申请日：2018-04-26
发明人： Lin Yang , Patrick Z. Dong , Charles Jin Young , Jason Dong , Wenhan Zhang , Baohua Sun
CPC分类号： G06K9/00288 , G06K9/00067 , G06K9/00375 , G06K9/4628 , G06K9/6256 , G06K9/6274 , G06K9/66 , G06K2209/01 , G06N3/0454 , G06N3/08 , G06N5/003
摘要： Image classification system contains a CNN based IC configured for extracting features out of input data by performing convolution operations using filter coefficients of ordered convolutional layers and a classifier IC configured for classifying the input data using reduced set of the extracted features based on a light-weight classifier. Light-weight classifier is derived by: training filter coefficients of the ordered convolutional layers using a dataset containing N labeled data, the trained filter coefficients are for the CNN based IC; outputting respective extracted features of the N labeled data after performing convolution operations of ordered convolutional layers using the trained filter coefficients, each labeled data contains X features; creating the reduced set of the extracted features by eliminating those of the X features that contain zeros in at least M of the N labeled data; and adjusting M until the light-weight classifier achieves satisfactory results using the reduced set.
-
8.发明授权公开(公告)号：US10452955B2
公开(公告)日：2019-10-22
申请号：US15871945
申请日：2018-01-15
发明人： Xiang Gao , Lin Yang , Wenhan Zhang
IPC分类号： G06K9/00 , G06K9/62 , G06K9/66 , G06F13/14 , G06E1/00 , G06K9/46 , G06N3/08 , G06N3/04 , G06T9/00
摘要： Methods of encoding image data for loading into an artificial intelligence (AI) integrated circuit are provided. The AI integrated circuit may have an embedded cellular neural network for implementing AI tasks based on the loaded image data. An encoding method may apply image splitting, principal component analysis (PCA) or a combination thereof to an input image to generate a plurality of output images. Each output image has a size smaller than the size of the input image. The method may load the output images into the AI chip, execute programming instructions contained in the AI chip to generate an image recognition result based on the at least one of the plurality of output images, and output the image recognition result. The encoding method also trains a convolution neural network (CNN) and loads the weights of the CNN into the AI integrated circuit for implementing the AI tasks.
-
公开(公告)号：US10402628B2
公开(公告)日：2019-09-03
申请号：US15963990
申请日：2018-04-26
发明人： Lin Yang , Patrick Z. Dong , Charles Jin Young , Jason Dong , Wenhan Zhang , Baohua Sun
摘要： Image classification system contains a CNN based IC configured for extracting features out of input data by performing convolution operations using filter coefficients of ordered convolutional layers and a classifier IC configured for classifying the input data using reduced set of the extracted features based on a light-weight classifier. Light-weight classifier is derived by: training filter coefficients of the ordered convolutional layers using a dataset containing N labeled data, the trained filter coefficients are for the CNN based IC; outputting respective extracted features of the N labeled data after performing convolution operations of ordered convolutional layers using the trained filter coefficients, each labeled data contains X features; creating the reduced set of the extracted features by eliminating those of the X features that contain zeros in at least M of the N labeled data; and adjusting M until the light-weight classifier achieves satisfactory results using the reduced set.
-
10.发明申请公开(公告)号：US20190220700A1
公开(公告)日：2019-07-18
申请号：US15871945
申请日：2018-01-15
发明人： Xiang Gao , Lin Yang , Wenhan Zhang
摘要： Methods of encoding image data for loading into an artificial intelligence (AI) integrated circuit are provided. The AI integrated circuit may have an embedded cellular neural network for implementing AI tasks based on the loaded image data. An encoding method may apply image splitting, principal component analysis (PCA) or a combination thereof to an input image to generate a plurality of output images. Each output image has a size smaller than the size of the input image. The method may load the output images into the AI chip, execute programming instructions contained in the AI chip to generate an image recognition result based on the at least one of the plurality of output images, and output the image recognition result. The encoding method also trains a convolution neural network (CNN) and loads the weights of the CNN into the AI integrated circuit for implementing the AI tasks.
-
-
-
-
-
-
-
-
-
搜索结果
18 条记录 (耗时 0.014 秒)
