Abstract: In the described examples, a compiled image scaler includes a set of machine executable instructions that generate a scaled image that is a scaled version of a source image with integer and bitwise operations. The compiled image scaler employs filtering to blend colors of adjacent pixels in the source image to generate the scaled image, and each filtering operation concurrently scales three color channels of a pixel in the source image.

Abstract: A method for training a perception network includes (a) perceiving first image-level data obtained from a first imaging device through the perception network to generate first prediction results, and training the perception network based on the first prediction results, (b) augmenting the first and second image-level data, respectively obtained from the first and a second imaging device, through a transfer network to generate first and second feature-level data, perceiving the first and the second feature-level data through the perception network to generate second prediction results, and training the transfer network based on the second prediction results, and (c) augmenting the first and the second image-level data through the transfer network to generate third feature-level data, perceiving the third feature-level data through the perception network to generate third prediction results, and retraining the perception network based on the third prediction results.

Abstract: A method and a system for joint data augmentation and classification learning, where an augmentation network learns to perform transformations and a classification network is trained. A set of labelled images is received. During an inner loop iteration, an augmentation network applies a transformation on a given labelled image of the set to obtain a transformed image. The classification network classifies the transformed image to obtain a predicted class, and a training loss is determined based on the predicted class and the respective label. The parameters of the classification network are updated based on the classification loss. During an outer loop iteration, the classification network classifies another labelled image of the set to obtain another predicted class, and a validation loss is determined based on the other predicted class and the respective label. The parameters of the augmentation network are updated based on the validation loss.

Abstract: Techniques of editing images using a digital image editing tool involve defining a local adjustment that includes, as separate masks, a spatial mask and a range mask. The spatial mask specifies a region of the image to be adjusted and the range mask specifies ranges of colors and tones to be adjusted independent of the region specified by the spatial mask. In applying these masks, when a user changes a setting in the digital image editing tool, the digital image editing tool weights the effect of the setting according to the values of the spatial mask and the range mask for each pixel of the image.

Abstract: A computer-implemented detection method includes, in response to inputting a first image including a region of one or more objects to a learned model, identifying a first entire image corresponding to entirety of a first object as a detection candidate, the learned model being generated by learning training data including an image corresponding to a part of an object and an entire image corresponding to entirety of the object, detecting an existing region of the first target object in the first image in accordance with a comparison between the identified first entire image and the region of the one or more target objects, and determining, based on a specific image obtained by invalidating the existing region in the first image, whether another target object is included in the first image.

Abstract: Techniques to mimic a target food item using artificial intelligence are disclosed. A formula generator is trained using ingredients and using recipes and, given a target food item, determines a formula that matches the given target food item. A flavor generator is trained using recipes and their associated flavor information and, given a formula, the flavor generator determines a flavor profile for the given formula. The flavor profile may be used to assist the formula generator in generating a subsequent formula. A recipe generator is trained using recipes and, given a formula, determines a cooking process for the given formula. A food item may be cooked according to a recipe, and feedback, including a flavor profile, may be provided for the cooked food item. The recipe and its feedback may be added to a training set for the flavor generator.

Abstract: Examples of techniques for adaptive model training are provided. According to one or more embodiments of the present invention, a computer-implemented method for adaptive model training includes generating, by a processing system, a training instance based at least in part on a plurality of images that match a contextual specification of a target visual domain. The method further includes extracting, by the processing system, objects from one of the plurality of images. The method further includes for each extracted object, generating, by the processing system, a plurality of machine learning model features and label recommendations for a user.

Abstract: Examples of techniques for adaptive object recognition for a target visual domain given a generic machine learning model are provided. According to one or more embodiments of the present invention, a computer-implemented method for adaptive object recognition for a target visual domain given a generic machine learning model includes creating, by a processing device, an adapted model and identifying classes of the target visual domain using the generic machine learning model. The method further includes creating, by the processing device, a domain-constrained machine learning model based at least in part on the generic machine learning model such that the domain-constrained machine learning model is restricted to recognize only the identified classes of the target visual domain. The method further includes computing, by the processing device, a recognition result based at least in part on combining predictions of the domain-constrained machine learning model and the adapted model.

Abstract: Automated discovery of the relative positioning of a network of cameras that view a physical environment. The automated discovery is based on comparing TimeLines for the cameras. The TimeLines are time-stamped data relating to the camera's view, for example a sequence of time stamps and corresponding images captured by a camera at those time stamps. In one approach, the relative positioning is represented by a proximity graph of nodes connected by edges. The nodes represent spaces in the physical environment, and each edge between two nodes represents a pathway between the spaces represented by the two nodes.

Abstract: Methods, apparatus, and articles of manufacture providing an efficient safety mechanism for signal processing hardware are disclosed. An example apparatus includes an input interface to receive an input signal; a hardware accelerator to process the input signal, the hardware accelerator including: unprotected memory to store non-critical data corresponding to the input signal; and protected memory to store critical data corresponding to the input signal; and an output interface to transmit the processed input signal.

Abstract: An apparatus and method for geometrically correcting an arbitrary shaped input frame and generating an undistorted output frame. The method includes capturing arbitrary shaped input images with multiple optical devices and processing the images, identifying redundant blocks and valid blocks in each of the images, allocating an output frame with an output frame size and dividing the output frame into regions shaped as a rectangle, programming the apparatus and disabling processing for invalid blocks in each of the regions, fetching data corresponding to each of the valid blocks and storing in an internal memory, interpolating data for each of the regions with stitching and composing the valid blocks for the output frame and displaying the output frame on a display module.

Abstract: A system and method for acquiring imaging data of an object. The system includes a beam splitter for directing a derivative sample beam towards the object and a derivative reference beam towards a reflective element; a detector for detecting at least one of the returned reference beam and the returned sample beam; a mode switching module for selecting an operating mode between an optical coherence tomography (OCT) mode and a hyperspectral imaging (HSI) mode; and a computing module for receiving the detection from the detector and the mode from the mode switching module, wherein in the OCT mode, the computing device generates OCT imaging data by determining an interference pattern produced by a superposition of the returned reference beam and the returned sample beam, and wherein in the HSI mode, the computing module generating hyperspectral imaging data by determining hyperspectral information of the returned sample beam.

Abstract: Aspects of the present disclosure describe systems, methods and structures employing time-space de-noising for distributed sensor.

Abstract: Examples of techniques for adaptive object recognition for a target visual domain given a generic machine learning model are provided. According to one or more embodiments of the present invention, a computer-implemented method for adaptive object recognition for a target visual domain given a generic machine learning model includes creating, by a processing device, an adapted model and identifying classes of the target visual domain using the generic machine learning model. The method further includes creating, by the processing device, a domain-constrained machine learning model based at least in part on the generic machine learning model such that the domain-constrained machine learning model is restricted to recognize only the identified classes of the target visual domain. The method further includes computing, by the processing device, a recognition result based at least in part on combining predictions of the domain-constrained machine learning model and the adapted model.

Abstract: In at least one aspect, a computer system is provided that includes at least one computer having software stored thereon that when executed causes the at least one computer to perform a method that includes the step or steps of: receiving at least one image that depicts a structure on real property, the structure having a first floor and at least one structural element depicted in the at least one image; dividing the image into a plurality of regions; predicting at least one bounding box for the at least one structural element in the at least one image; predicting that the bounding box contains a structural element of interest; estimating a first floor elevation or height of the structure on the real property based on the structural element of interest; and storing the estimated first floor elevation or height in at least one database.

Abstract: A monitoring system includes a communications circuit and an analysis circuit where the communications circuit is configured to receive an image of a user from a social media system and receive orthodontic treatment plan data associated with the user from an aligner fabrication computer system. The analysis circuit is configured to determine an actual oral condition of the user at a time when the image of the user was taken, determine an expected oral condition of the user at the time when the image of the user was taken based on the orthodontic treatment plan data, compare the actual oral condition and the expected oral condition, and generate an output based on the comparison. The communications system communications the output to at least one of the user device and the aligner fabrication computer system.

Abstract: An object position and/or orientation can be determined based on image data and object contact points. Image data can be captured representing an object, such as a vehicle. Vehicle contact points can be identified in the image data representing wheel contacts with the ground. For an individual vehicle contact point (e.g., a left-front wheel of the second vehicle), a ray can be determined that emanates from the image sensor and passes through the vehicle contact point. To determine a location and velocity of the vehicle, the ray can be unprojected onto a three-dimensional surface mesh, and an intersection point between the ray and the three-dimensional surface mesh can be used as an initial estimate for the projected location of the vehicle contact point in the world. The estimated location can be adjusted based on various cost functions to optimize an accuracy of the locations of the estimated vehicle contact points.

Abstract: A computer implemented method for determining survey sampling parameters for environmental DNA (eDNA) detection comprises receiving a species selection identifying selected species and receiving environmental specifications for an environment to be tested for presence of the species. A sampling plan is generated using the environmental specifications and the species selection, and detectability prediction(s) are generated using the environmental specifications, the species selection and the current sampling plan to predict whether the selected species is detectable in the environment according to the current sampling plan. Where at least one respective selected species is undetectable according to the current sampling plan, the process iterates, with each subsequent iteration incorporating an increase in the total volume to be sampled, until either every respective selected species is detectable according to the then-current sampling plan or an iteration stop limit is reached.

Abstract: Identification of augmented features based on a Bayesian analysis of a text document is disclosed. One example is a system including a document processing module, a feature processing module, and a feature generation module. The document processing module receives a text document via a processor. The feature processing module automatically identifies, based on a Bayesian analysis of the text document, a plurality of augmented features in the text document, the plurality of augmented features including at least one of local, sectional, and document-level features of the text document, and extracts, via the processor, the identified plurality of augmented features from the text document. The feature generation module generates, via the processor, a feature representation of the text document based on the extracted plurality of augmented features.

Abstract: A communication device is provided for a motor vehicle that works with traffic-relevant information provided through light signals. The communication device comprises a primary imaging optical system to create the input image from the received scenic image, which may comprise a light car2car signal and/or a light car2X signal including traffic-relevant information. It further comprises a DMD array of mirrors for receiving the input image and transmitting it to a primary optical sensor adapted to convert the input optical image to an electric signal for processing, to thereby find out whether and what places of the input image sources of the signals are found in.