A camera system captures images from a set of cameras to generate binocular panoramic views of an environment. The cameras are oriented in the camera system to maximize the minimum number of cameras viewing a set of randomized test points. To calibrate the system, matching features between images are identified and used to estimate three-dimensional points external to the camera system. Calibration parameters are modified to improve the three-dimensional point estimates. When images are captured, a pipeline generates a depth map for each camera using reprojected views from adjacent cameras and an image pyramid that includes individual pixel depth refinement and filtering between levels of the pyramid. The images may be used generate views of the environment from different perspectives (relative to the image capture location) by generating depth surfaces corresponding to the depth maps and blending the depth surfaces.

Methods, systems and computer-readable media for extracting data from a document. One method includes receiving a document in a text format and assigning a first signature to the document. The method also includes matching the first signature to a second signature of a template and extracting data a user desires to have extracted from the document, wherein instructions for locating the data are stored within the template.

Methods, systems, and devices for object recognition are described. A device may generate a subspace based at least in part on a set of representative feature vectors for an object. The device may obtain an array of pixels representing an image. The device may determine a probe feature vector for the image by applying a convolutional operation to the array of pixels. The device may create a reconstructed feature vector in the subspace based at least in part on the set of representative feature vectors and the probe feature vector. The device may compare the reconstructed feature vector and the probe feature vector and recognize the object in the image based at least in part on the comparison. For example, the described techniques may support pose invariant facial recognition or other such object recognition applications.

A method includes passing an original text document through distortion filter generators to generate a training dataset that includes distorted text documents. Each distortion filter generator is configured to distort words or letters of words in phrases of text of a facsimile image in a respective unique manner. A neural network model is trained to recognize each respective distortion and match each respective distortion with each respective distortion filter generator based on the training dataset and the original text document. Image data of one facsimile having at least one text distortion is received and inputted to the trained neural network model. The output of the trained neural network model is coupled to an input of an optical character recognition (OCR) engine. The trained neural network model and the OCR engine convert the received image data of the incoming facsimile corrected for the at least one text distortion to machine-encoded text.

Introduced here is a machine learning related technique for supplying an observed model additional training data based upon previously received training data. To determine textual content of a character string based on a digital image that includes a handwritten version of the character string a substantial amount of training data is used. The character string can include one or more characters, and the characters can include any of letters, numerals, punctuation marks, symbols, spaces, etc. Disclosed herein is a technique to determine variations between different images of matching known character strings and substitute those variations into the images in order to create more images with the same known character string.

A determination method for determining the structure of a convolutional neural network includes acquiring N filters having the weights trained using a training image group as the initial values, where N is a natural number greater than or equal to 1, and increasing the number of the filters from N to M, where M is a natural number greater than or equal to 2 and is greater than N, by adding a filter obtained by performing a transformation used in image processing fields on at least one of the N filters.

Methods, systems, and computer programs are presented for a classifier that recognizes images when at least one class includes just a few training samples. One method includes identifying a training set containing samples, each sample associated with a class from many classes. Further, the method divides the training set into a base and a novel set based on the number of samples in each class, trains a first classifier with the base set, and trains a second classifier using the training set. The second classifier is trained with promotion of the novel set and based on minimizing a loss function that comprises a first term and a second term, the first term associated with a first summation for the samples of the training set, the second term associated with a second summation for the samples of the novel set. Further, the method classifies an item with the trained second classifier.

A camera system captures images from a set of cameras to generate binocular panoramic views of an environment. The cameras are oriented in the camera system to maximize the minimum number of cameras viewing a set of randomized test points. To calibrate the system, matching features between images are identified and used to estimate three-dimensional points external to the camera system. Calibration parameters are modified to improve the three-dimensional point estimates. When images are captured, a pipeline generates a depth map for each camera using reprojected views from adjacent cameras and an image pyramid that includes individual pixel depth refinement and filtering between levels of the pyramid. The images may be used generate views of the environment from different perspectives (relative to the image capture location) by generating depth surfaces corresponding to the depth maps and blending the depth surfaces.

Introduced here is a machine learning related technique for supplying an observed model additional training data based upon previously received training data. To determine textual content of a character string based on a digital image that includes a handwritten version of the character string a substantial amount of training data is used. The character string can include one or more characters, and the characters can include any of letters, numerals, punctuation marks, symbols, spaces, etc. Disclosed herein is a technique to determine variations between different images of matching known character strings and substitute those variations into the images in order to create more images with the same known character string.

Techniques are disclosed for generating a sequence of symbols based on input data for a neuro-linguistic model. The model may be used by a behavior recognition system to analyze the input data. A mapper component of a neuro-linguistic module in the behavior recognition system receives one or more normalized vectors generated from the input data. The mapper component generates one or more clusters based on a statistical distribution of the normalized vectors. The mapper component evaluates statistics and identifies statistically relevant clusters. The mapper component assigns a distinct symbol to each of the identified clusters.