Abstract: A video content recognition method is performed by a computer device, the method including: obtaining an image feature corresponding to a video frame set extracted from a target video; dividing the image feature into a plurality of image sub-features according to a preset sequence, and each image sub-feature having a corresponding channel; choosing, from the image sub-features based on the preset sequence, a current image sub-feature; image sub-feature fusing the current image sub-feature and a convolution processing result of a previous image sub-feature into a fused image sub-feature, and performing convolution processing on the fused image sub-feature, to obtain a convolved image sub-feature corresponding to the current image sub-feature; splicing a plurality of convolved image sub-features corresponding to the plurality of channels of the convolved image sub-feature, to obtain a spliced image feature; and determining video content corresponding to the target video based on the spliced image feature.

Abstract: Systems and methods for generating training data using synthesized input images for training a machine learning model for image enhancement in accordance with embodiments of the invention are described. The system may access a target image (e.g., captured by an imaging device), and generate an input image that corresponds to the target image. The input image and target image may then be used (e.g., as part of a training data set) to train the machine learning model. For example, a generated input image corresponding to a target image may represent a version of the target image as if it were captured in low light. The target image may be a target illuminated output to be generated by enhancing the input image. The input image and target image may be used to train a machine learning model to enhance images such as those captured in low light.

Abstract: The present invention generally relates to assigning and reporting the cleanliness of objects and areas, and particularly relates to utilization of object and area cleanliness states as determined by the opposing processes of cleaning and dirtying detected through various methods to provide an indication of the state of cleanliness potentially utilized to alter the process of cleaning and/or dirtying to reach a desired state of cleanliness. Detection of the cleaning and dirtying operations can be performed automatically through image processing and behavior detection of still images/video indicating the activity taking place in the area of interest and time. The state of cleanliness can then be reported to interested parties as textual reports and/or augmented reality overlays on still images/video.

Abstract: Advancing beyond Interpretability and explainability approaches that may uncover what a Deep Neural Network (DNN) models, i.e., what each node (cell) in the network represents and what images are most likely to activate the model provide a mimicked type of learning of generalization applicable to previously unseen samples. The approach provides an ability to detect and circumvent adversarial attacks, with self-verification and trust-building structural modeling. Computing systems may now define what it means to learn in deep networks, and how to use this knowledge for a multitude of practical applications.

Abstract: A state determination device includes: an extraction unit that extracts a face region representing a region corresponding to a face from each of a plurality of frames successively obtained by capturing images of a user's face, extracts face feature points representing parts of the face from the face region, calculates a face feature value extraction region as a region where a change occurs in the face region when the user is in an unawakened state based on the face feature points, and extracts a face feature value as a feature value from the face feature value extraction region; a state determination unit that determines whether the user is in the unawakened state or not based on the face feature value in each of the plurality of frames and previously generated determination information; and an output unit that outputs a determination result.

Abstract: A computer-implemented method for explaining an image classifier, the method comprising: receiving an initial image, the initial image having been wrongly classified by the image classifier; receiving an initial gradient of a function executed by the image classifier generated while classifying the initial image, the function being indicative of a probability for the initial image to belong to an initial class; converting the initial image into a latent vector, the latent vector being a representation of the initial image in a latent space; generating a plurality of perturbation vectors using the initial gradient of the function executed by the image classifier; combining the latent vector with each one of the plurality of perturbation vectors, thereby obtaining a plurality of modified vectors; for each one of the plurality of modified vectors, reconstructing a respective image, thereby obtaining a plurality of reconstructed images; transmitting the reconstructed images to the image classifier; for each one o

Abstract: A computer-implemented method for detecting a target object on a document page that includes detecting a sample target area on a sample document page, generating an image by overlapping a plurality of sample document pages with one another, and detecting one or more cells within the sample target area on the image. The sample target area includes a sample target object. The method further includes extracting one or more informational features from each of the one or more cells. The one or more informational features define characteristics of a corresponding cell of the one or more cells. A machine learning model is trained using the one or more informational features extracted from each of the one or more cells, to detect the sample target object. A target object on a document page is detected using the trained machine learning model.

Abstract: Techniques for encoding or decoding digital video or pictures include acquiring a video bitstream that includes an encoded video image that is a two-dimensional image comprising multiple regions of a panoramic image in three-dimensional coordinates, extracting neighboring information for the multiple regions, performing, using the neighboring information, post-processing of a video image decoded from the video bitstream, and generating a display image from the video image after the post-processing.

Abstract: An apparatus comprises a processing device configured to receive a service request for a given computing device, the service request comprising a given image of a given computing device, to generate a given image embedding for the given image utilizing a machine learning model, and to determine similarity measures between the given image embedding and other image embeddings for other images of computing devices utilizing an angular similarity metric. The processing device is configured to identify whether the given image exhibits at least a threshold level of similarity to at least one other image based at least in part on the determined similarity measures, to classify the given image as potentially fraudulent responsive to identifying the given image as exhibiting at least the threshold level of similarity to at least one other image, and to modify processing of the service request responsive to classifying the given image as potentially fraudulent.

Abstract: In some aspects, a debris detector includes one or more processors and a memory storage device to store instructions executable by the one or more processors to: receive a frame captured by a camera, receive a subsequent frame captured by the camera, compare the frame to the subsequent frame to determine a difference, and based on determining that the difference is greater than a threshold, send a notification indicating that debris has been detected. The notification may include location data identifying a location where the debris is located. Receiving the notification may cause a robot vacuum to initiate cleaning the location identified in the notification.

Abstract: Disclosed are an information processing method and a terminal device. The method comprises: acquiring first information, wherein the first information is information to be processed by a terminal device; calling an operation instruction in a calculation apparatus to calculate the first information so as to obtain second information; and outputting the second information. By means of the examples in the present disclosure, a calculation apparatus of a terminal device can be used to call an operation instruction to process first information, so as to output second information of a target desired by a user, thereby improving the information processing efficiency. The present technical solution has advantages of a fast computation speed and high efficiency.

Abstract: A product count of a number of physical products within a physical grouping of a plurality of the physical products is determined based on an image of the physical grouping. A cycle counter generates image coordinates for visible product surfaces within the image. The cycle counter generates three-dimensional virtual base coordinates for the expected locations of the plurality of the physical products within the physical grouping. The cycle counter determines the actual locations of the visible product surfaces within three-dimensional space based on a comparison of the image coordinates and the virtual base coordinates. The cycle counter determines the product count of the number of physical products within the physical grouping based on the actual locations.

Abstract: An image transformation method includes: obtaining identification information of an original image; converting the identification information to an identification image; performing frequency domain transformation on the original image to obtain a pixel matrix of the original image in a frequency domain space; performing matrix decomposition on the pixel matrix to obtain an image brightness matrix; converting pixel values of corresponding pixels in the image brightness matrix based on pixel values of pixels in the identification image to obtain a converted brightness matrix; and performing inverse frequency domain transformation on the converted brightness matrix to obtain a transformed image including invisible identification information, and adding the invisible identification information to the original image.

Abstract: This disclosure describes one or more embodiments of systems, non-transitory computer-readable media, and methods that can learn or identify a learned-initialization-latent vector for an initialization digital image and reconstruct a target digital image using an image-generating-neural network based on a modified version of the learned-initialization-latent vector. For example, the disclosed systems learn a learned-initialization-latent vector from an initialization image utilizing a high number (e.g., thousands) of learning iterations on an image-generating-neural network (e.g., a GAN). Then, the disclosed systems can modify the learned-initialization-latent vector (of the initialization image) to generate modified or reconstructed versions of target images using the image-generating-neural network.

Abstract: A method and apparatus for generating an image and for training an artificial neural network to generate an image are provided. The method of generating an image, including receiving input data comprising conditional information and image information, generating a synthesized image by applying the input data to an image generation neural network configured to maintain geometric information of the image information and to transform the remaining image information based on the conditional information, and outputting the synthesized image.

Abstract: A system of enhancing biometric analysis matching utilizes an image sensor, such as a digital camera, to capture an image of a face of a person. The system may perform image enhancement, such as edge and contrast enhancement, prior to performing face matching. The enhancement may be localized to a given image region based on determined region illumination. The system may perform image processing and analysis comprising face detection, alignment, feature extraction, and recognition. A biometric recognition confidence indicator may be generated using the results of the image enhancement and analysis. At least partly in response to the biometric recognition confidence indicator falling below a threshold enhancing recognition confidence using an image of visual indicia captured using the image sensor.

Abstract: A method and system to determine the position of a moveable platform relative to an object is disclosed. The method can include storing one or more synthetic models each trained by one of the one or more synthetic model datasets corresponding to one or more objects in a database; capturing an image of the object by one or more sensors associated with the moveable platform; identifying the object by comparing the captured image of the object to the one or more synthetic model datasets; generating a first model output using a first synthetic model of the one or more synthetic models, the first model output including a first relative coordinate position and a first spatial orientation of the moveable platform; and generating a platform coordinate output and a platform spatial orientation output of the moveable platform at the first position based on the first model output.

Abstract: The present invention provides a method for estimating the permeability of rock from a digital image of the rock. A three-dimensional image of a rock is obtained and segmented, and an image permeability is determined from the segmented image of the rock. Permeability correction factors are obtained from the segmented image and from a non-wetting liquid capillary pressure curve derived from the segmented image, and the permeability correction parameters are applied to the image permeability to obtain a corrected image permeability of the rock.

Abstract: A tamper-evident glass and methods of using the tamper-evident glass are provided herein. The tamper-evident glass comprising at least a first plurality of tamper-evident elements, which may be unique or near unique to each manufacturing of tam per-evident glass. The initial state of tamper-evident glass and positioning of the plurality of tamper-evident elements may be mapped through digital scanning to create a digital file of the current state of the tamper-evident glass. The tamper-evident glass may be scanned again to create a new digital file for comparison and to determine if a tampering event has occurred. The tamper-evident elements may include any combination of inorganic fluorescent material, crystals grown through nucleation and crystallization, or added colorants.

Abstract: Various disclosed embodiments are directed to inpainting one or more portions of a target image based on merging (or selecting) one or more portions of a warped image with (or from) one or more portions of an inpainting candidate (e.g., via a learning model). This, among other functionality described herein, resolves the inaccuracies of existing image inpainting technologies.