Abstract: An image processing apparatus comprising: one or more memories storing instructions; and one or more processors executing the instructions to: execute first setting processing for setting a reduced region in an image; execute second setting processing for setting an estimation region in the image based on the reduced region; execute estimation processing for estimating a noise characteristic of the estimation region; and execute noise reduction processing that reduces noise in the reduced region of the image according to a parameter based on the noise characteristic, wherein in the second setting processing, when a surface area of the reduced region is smaller than a threshold surface area, the estimation region is set to be at least as large as the threshold surface area.

Abstract: A posture comparison method, an electronic device, and a computer-readable storage medium are provided. The method includes: obtaining M first image frames of a first video stream within a specific time interval, and obtaining N second image frames of a second video stream within the specific time interval; determining K first joint points corresponding to K specified joints in each first image frame, and determining K second joint points corresponding to the K specified joints in each second image frame; obtaining an m-th first image frame among the M first image frames and an n-th second image frame among the N second image frames; and determining a difference between each first joint point and each corresponding second joint point in the m-th first image frame and the n-th second image frame, and determining a posture difference degree and providing a posture correction recommendation corresponding to the specific time interval accordingly.

Abstract: Provided in the present disclosure is a model training method for video quality assessment, including: acquiring training video data, where the training video data includes reference video data and distorted video data; determining a mean opinion score (MOS) of each piece of training video data; and training a preset initial video quality assessment model according to the training video data and the MOS thereof, until a convergence condition is met, so as to obtain a final video quality assessment model. Further provided in the present disclosure are a video quality assessment method and apparatus, a device and a medium.

Abstract: Some methods are described herein for tracking one or more objects. In some examples, the method is performed at a computer system. In some examples, the method includes receiving a first set of data corresponding to the object via a first modality and a second set of data corresponding to the object via a second modality; after receiving the first set of data corresponding to the object and the second set of data corresponding to the object, receiving a second set of image data representing the field of view of the camera, wherein the second set of image data does not include data representative of the object; and after receiving the second set of image data, predicting a position of the object using at least the first set of data corresponding to the object and the second set of data corresponding to the object.

Abstract: A method for detecting color moiré in an image, including: obtaining image data corresponding to an input image, wherein the image data comprises chrominance channel data and luminance channel data; applying a frequency transform to the image data to obtain transformed image data; based on the transformed image data, selecting at least one lowest frequency component in the chrominance channel data, and at least one lowest frequency component in the luminance channel data; determining a first magnitude corresponding to the at least one lowest frequency component in the chrominance channel data, and a second magnitude corresponding to the at least one lowest frequency component in the luminance channel data; determining a moiré grade by calculating a ratio of the first magnitude and the second magnitude; and detecting a color moiré artifact in the input image based on the moiré grade.

Abstract: A detection method for detecting fault of insulator discharge based on image recognition includes the following steps: building an ultraviolet image dataset for an electrical equipment, building a neural network model for insulator discharge detection based on YOLOV5-T, the neural network model includes a CBS module, a CST module, a Concat module, a 1D-attention module, an Upsample module and a Detect module, training the neural network model by utilizing the ultraviolet image dataset for the electrical equipment, to obtain a trained neural network model, and inputting images to be recognized into the trained neural network model, to obtain fault detection results. A detection device is configured to implement the detection method.

Abstract: A technique is disclosed for analyzing and displaying the extent to which the images and structures inferred by a physically seeded multiscale network correspond to genuine resolution improvement and the extent to which they correspond to the hallucination of realistic looking structures. A selected reconstruction generated using a trained neural network is compared against a baseline representation (e.g., a baseline reconstruction) by calculating image similarity metrics between progressing spatially filtered versions of selected reconstruction. As the amount of spatial filtering increases, at some point the image similarity metrics will reach an extrema (e.g., a lowest distance between the images). At that extrema, the parameter(s) of the spatial filter can be used to identify a resolution score (e.g., a length scale) associated with that extrema. The resolution score is indicative of an amount of resolution recovery associated with the trained neural network with respect to the baseline.

Abstract: A method for determining an image quality verification (IQV) metric for false color artifacts in chromatic areas of digital images includes receiving a digital image that includes flat pure color patches and mixed color texture patches; processing the digital image in an image signal processing chain such that a red/green/blue image is produced; transforming the red/green/blue image into an LAB color space image; calculating, for each mixed color pixel in a mixed color texture patch, a chromatic distance in the LAB color space from the mixed color pixel in the mixed color texture patch to a center of each flat pure color patch; performing directional adjustments on each of the chromatic distances; and defining, for each mixed color patch, the IQV metric for the for the mixed color patch as a function of the distance metrics of each mixed color texture pixel in the mixed color patch.

Abstract: In a correction mode, a correction image is displayed on an extension display, and a specific region is displayed on the extension display. In the correction mode, a display style control unit performs at least one of changing a display style of the correction image or changing a display style of the specific region in accordance with reliability related to calculation of the oxygen saturation.

Abstract: An image processing apparatus includes a memory storing instructions, and a processor configured to execute the instructions to acquire information about an optical system included in an image pickup apparatus configured to capture an image, acquire a change amount in the information about the optical system in a peripheral area of a target pixel in the image based on the information about the optical system, and acquire, based on the change amount, information about recognition accuracy in the peripheral area by a recognition apparatus configured to perform image recognition.

Abstract: 1.

Abstract: There is provided a method for training a caption rephrasing model to allow edge-case data to be generated. The method includes steps of: (a) in response to acquiring at least one image caption, inputting, by a learning device, the image caption to the caption rephrasing model to generate a first rephrased caption to an n-th rephrased caption; (b) inputting, by the learning device, each of the first rephrased caption to the n-th rephrased caption to an image generating model, to thereby generate a first generated image and a first pseudo-label to an n-th generated image and an n-th pseudo-label, and generating, by the learning device, a first identification loss to an n-th identification; and (c) fine-tuning, by the learning device, the caption rephrasing model through a preference learning using one of (i) a preference data set and (ii) a group normalized reward.

Abstract: A computer device segments a first sample region to obtain a candidate image region set that includes a plurality of candidate image regions. For each of the candidate image regions, the device obtains a first relationship degree corresponding to each candidate image region and obtains a second relationship degree corresponding to the candidate image region. The device obtains a relationship degree change value based on the first relationship degree and the second relationship degree. The device selects, from the plurality of candidate image regions, a first candidate image region as a target image region in accordance with a determination that the first candidate image region satisfies a condition in the relationship degree change value. The device performs model training based on the target image region to obtain a target detection model.

Abstract: Generating intelligent drawings includes obtaining a current set of design drawings for a physical plant comprising a plurality of components, generating rasterized drawings from the current set of design drawings, and applying an object detection model to the rasterized drawings to extract a set of shapes corresponding to components in the drawings. A trained library is applied to the extracted shapes to obtain a classification of components in the drawings, and the classification is propagated to instances of the components in the current set of drawings.

Abstract: Various embodiments relate generally to computer vision and automation to autonomously survey an agricultural environment. In some examples, a system can receive sensor data of a geographic boundary from one or more sensors onboard of the agricultural observation system, detect a plurality of external dynamic objects, a plurality of external static objects, one or more agricultural objects, or a combination thereof from the sensor data, determine a position of a component of the agricultural observation system relative to the plurality of external dynamic objects, the plurality of external static objects, the one or more agricultural objects, or a combination thereof, and generate a local map, in real time, via one or more processing units.

Abstract: An information processing apparatus comprises a first inference unit configured to infer an attribute of an object in an input image, and a weight relating to the attribute of each region in the input image, and a second inference unit configured to infer an attribute of the object, based on an attribute and a weight that are inferred by the first inference unit.

Abstract: Apparatuses, systems, and techniques to scale textured images using a Fourier transform in conjunction with one or more neural networks. In at least one embodiment, a neural network generates an expanded image from an input image by applying a Fourier transform to one or more feature maps generated by said neural network and up-scaling one or more resulting frequency domain feature maps before generating an expanded output image based on up-scaled feature maps.

Abstract: The present disclosure relates to systems and methods for imaging. The method may include obtaining a first image and topology data of an object, wherein the topology data may include first topology data and second topology data, and the first topology data may correspond to the second topology data. The method may also include determining a base material density image corresponding to the first image, determining a difference in the topology data based on the first topology data and the second topology data, and determining a second image corresponding to the first image based on the first image, the base material density image, and the difference in the topology data.

Abstract: This document describes an accelerator circuitry for facilitating acceleration of non-maximum suppression (NMS) for detection of objections within an image.

Abstract: Disclosed herein is a system and method for reducing false positives in object detection frameworks. A human reviews objects detected by the object detection framework and indicates whether the object is an object of a class for which the object detection framework is trained to detect. When an indication of a false positive been received, a feature representation of displayed object is stored in a gallery. During an inference or testing phase, the gallery is searched for a feature representation matching the feature representation of the detected objects, and, if a match is found, the detected object is deemed to be a false positive and is not identified as an object for which the object detection framework has been trained to detect.