Abstract: The present disclosure relates to devices, apparatus and methods of improving the accuracy of image-based assay, that uses imaging system having uncertainties or deviations (imperfection) compared with an ideal imaging system. One aspect of the present invention is to add the monitoring marks on the sample holder, with at least one of their geometric and/optical properties of the monitoring marks under predetermined and known, and taking images of the sample with the monitoring marks, and train a machine learning model using the images with the monitoring mark.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for image rendering. In one aspect, a method comprises receiving a plurality of observations characterizing a particular scene, each observation comprising an image of the particular scene and data identifying a location of a camera that captured the image. In another aspect, the method comprises receiving a plurality of observations characterizing a particular video, each observation comprising a video frame from the particular video and data identifying a time stamp of the video frame in the particular video. In yet another aspect, the method comprises receiving a plurality of observations characterizing a particular image, each observation comprising a crop of the particular image and data characterizing the crop of the particular image. The method processes each of the plurality of observations using an observation neural network to determine a numeric representation as output.

Abstract: Aspects of the disclosure provide for mechanisms for identification of objects in images using neural networks. A method of the disclosure includes: obtaining an image, representing each element of a plurality of elements of the image via an input vector of a plurality of input vectors, each input vector having one or more parameters pertaining to visual appearance of a respective element of the image, providing the plurality of input vectors to a first subnetwork of a neural network to obtain a plurality of output vectors, wherein each of the plurality of output vectors is associated with an element of the image, identifying, based on the plurality of output vectors, a sub-plurality of elements of the image as belonging to the image of the object, and determining, based on locations of the sub-plurality of elements, a location of an image of an object within the image.

Abstract: The present invention relates to the determination of damage to portions of a vehicle. More particularly, the present invention relates to determining whether each part of a vehicle should be classified as damaged or undamaged and optionally the severity of the damage to each part of the damaged vehicle including preserving the quality of the input images of the damage to the vehicle. Aspects and/or embodiments seek to provide a computer-implemented method for determining damage states of each part of a damaged vehicle, indicating whether each part of the vehicle is damaged or undamaged and optionally the severity of the damage to each part of the damaged vehicle, using images of the damage to the vehicle and trained models to assess the damage indicated in the images of the damaged vehicle, including preserving the quality and/or resolution of the images of the damaged vehicle.

Abstract: According to one embodiment, an image processing apparatus includes a buffer and processing circuitry. The buffer stores first and second images capturing an object. The circuitry calculates at least one of a first distance to the object in the first image and a second distance to the object in the second image by using a correction parameter for correcting at least one of influences caused by ambient light, a reflection characteristic of the object, or a color of the object, calculates three-dimensional coordinates of the object on a relative scale by using the first and second images, and calculates three-dimensional coordinates of the object on a real scale based on at least one of the first and second distances, and the three-dimensional coordinates of the object on the relative scale.

Abstract: The recognition system for security check and control method thereof. The recognition system for security check is integrated with a reinforcement learning algorithm and an attention region proposal network. The recognition system for security check comprises the following modules: an object feature extraction module (1); a dangerous item region segmentation module (2); a preliminary classification module (3); a preliminary classification result determination module (4); and a fine-grained recognition module (5). In the invention, optimization of a dangerous item region segmentation module and provision of a fine-grained recognition module greatly improve accuracy and efficiency of security check, shorten the duration of security check, alleviate congestion, save labor, and reduce pressure on security check personnel.

Abstract: In measuring a dimension of an object to be measured W made of a single material, a plurality of transmission images of the object to be measured W are obtained by using an X-ray CT apparatus, and then respective projection images are generated. The projection images are registered with CAD data used in designing the object to be measured W. The dimension of the object to be measured W is calculated by using a relationship between the registered CAD data and projection images. In such a manner, high-precision dimension measurement is achieved by using several tens of projection images and design information without performing CT reconstruction.

Abstract: A system and a method are disclosed for a structured-light system to estimate depth in an image. An image is received in which the image is of a scene onto which a reference light pattern has been projected. The projection of the reference light pattern includes a predetermined number of particular sub-patterns. A patch of the received image and a sub-pattern of the reference light pattern are matched based on either a hardcode template matching technique or a probability that the patch corresponds to the sub-pattern. If a lookup table is used, the table may be a probability matrix, may contain precomputed correlations scores or may contain precomputed class IDs. An estimate of depth of the patch is determined based on a disparity between the patch and the sub-pattern.

Abstract: Methods and systems for learnable defect detection for semiconductor applications are provided. One system includes a deep metric learning defect detection model configured for projecting a test image for a specimen and a corresponding reference image into latent space, determining a distance in the latent space between one or more different portions of the test image and corresponding portion(s) of the corresponding reference image, and detecting defects in the one or more different portions of the test image based on the determined distances. Another system includes a learnable low-rank reference image generator configured for removing noise from one or more test images for a specimen thereby generating one or more reference images corresponding to the one or more test images.

Abstract: Embodiments of the present invention provide an improvement to conventional machine model training techniques by providing an innovative system, method and computer program product for the generation of synthetic data using an iterative process that incorporates multiple machine learning models and neural network approaches. A collaborative system for receiving data and continuously analyzing the data to determine emerging patterns is provided. The proposed invention involves generating synthetic data clusters to be stored and used for retraining the main model as well as other models. In addition, the invention includes using one or more (subset) of the synthetic data clusters to train or retrain machine learning models, developing and training machine learning models that are trained with emerging synthetic data clusters, and ensembling machine learning models trained with emerging synthetic data clusters.

Abstract: Methods and apparatus are disclosed for implementing data augmentation within a storage controller of a data storage device based on machine learning data read from a non-volatile memory (NVM) array of a memory die. Some particular aspects relate to configuring the storage controller to generate augmented versions of training images for use in training a Deep Learning Accelerator of an image recognition system by rotating, translating, skewing, cropping, etc., a set of initial training images obtained from a host device and stored in the NVM array. Other aspects relate to controlling components of the memory die to generate noise-augmented images by, for example, storing and then reading training images from worn regions of the NVM array to inject noise into the images. Data augmentation based on data read from multiple memory dies is also described, such as image data spread across multiple NVM arrays or multiple memory dies.

Abstract: In a data driven object recognition system and object recognition method, a connection relationship between an object feature extraction unit and a plurality of task-specific identification units is stored in a connection switch according to a type of task. The connection relationship is changed based on the connection information to suppress the amount of labeling in constructing a learning data set.

Abstract: Embodiments of the present disclosure disclose a method and apparatus for processing an image. A specific embodiment of the method includes: acquiring a feature map of a target image, where the target image contains a target object; determining a local feature map of a target size in the feature map; combining features of different channels in the local feature map to obtain a local texture feature map; and obtaining location information of the target object based on the local texture feature map.

Abstract: The disclosure relates to systems, methods and programs for geometrically constrained, unsupervised training of convolutional autoencoders on unlabeled images for extracting eye landmarks. Disclosed systems for unsupervised deep learning of gaze estimation in eyes' image data are implementable in a computerized system. Disclosed methods include capturing an unlabeled image comprising the eye region of a user; and training a plurality of convolutional autoencoders on the unlabeled image comprising the eye region of a user using an initial geometrically regularized loss function to determine a plurality of eye landmarks.

Abstract: Methods and apparatus are disclosed for implementing machine learning data augmentation within the die of a non-volatile memory (NVM) apparatus using on-chip circuit components formed on or within the die. Some particular aspects relate to configuring under-the-array or next-to-the-array components of the die to generate augmented versions of images for use in training a Deep Learning Accelerator of an image recognition system by rotating, translating, skewing, cropping, etc., a set of initial training images obtained from a host device. Other aspects relate to configuring under-the-array or next-to-the-array components of the die to generate noise-augmented images by, for example, storing and then reading training images from worn regions of a NAND array to inject noise into the images.

Abstract: A method, storage media and neural network for rebuilding biometric feature is provided. The method includes inputting the partial texture image obtained to the neural network and outputting a predictive value of an entire texture image that is output by the neural network. The above technical solution is via the neural network used to process images and the neural network includes the feature value layer. A plurality of the partial texture images is converted to the feature values at the technical level, and the composite calculation of a plurality of partial texture images is avoided on the application level. Because the entire texture image is not synthesized in the end, data leakage and theft are avoided. Thus, the security of the method for analyzing texture image is improved.

Abstract: An image processing apparatus includes: an object detection section that performs convolution computation on an input image based on a captured image obtained by capturing the image with a camera, and that detects an object; a feature map validation section that performs feature map validation validating a likelihood that the input image contains the object on the basis of a feature map obtained by the convolution computation; a time series validation section that performs time series validation validating a result of the feature map validation performed by the feature map validation section in time series; and a detection result correction section that corrects a detection result about the object output by the object detection section on the basis of a result of the time series validation performed by the time series validation section.

Abstract: An information processing apparatus includes a first estimation unit configured to estimate, for each of a plurality of images successive in time series, the number of objects existing in each of a plurality of set regions, and a second estimation unit configured to estimate a flow of the objects existing in each of the plurality of regions based on a result of the estimation for each of the plurality of images by the first estimation unit.

Abstract: An apparatus receives, at a discriminator within a generative adversarial network, first generation data from a first generator within the generative adversarial network, where the first generator has performed learning using a first data group. The apparatus receives, at the discriminator, a second data group, and performs learning of a second generator based on the first generation data and the second data group where the first generation data is handled as false data by the discriminator.

Abstract: The present disclosure describes methods, devices, and storage medium for generating a time-lapse photography video with a neural network model. The method includes obtaining a training sample. The training sample includes a training video and an image set. The method includes obtaining through training according to the training sample, a neural network model to satisfy a training ending condition, the neural network model comprising a basic network and an optimization network, by using the image set as an input to the basic network, the basic network being a first generative adversarial network for performing content modeling, generating a basic time-lapse photography video as an output of the basic network, using the basic time-lapse photography video as an input to the optimization network, the optimization network being a second generative adversarial network for performing motion state modeling, and generating an optimized time-lapse photography video as an output of the optimization network.