Abstract: A system and storage medium for alerting crop disease or pest infestation includes satellite-based and aerial remote imaging devices for capturing RGB images of the Earth and attach Earth location and date information to each RGB image. Machine learning processing circuitry includes a stem module to convert the RGB images into a high-dimensional feature map, and a multi-scale feature extractor including consecutive branches to learn representations at different resolutions and semantic levels. Each branch includes adaptive inverted blocks and multi-scale channel attention module. Each adaptive inverted block includes a spatial spectral attention block configured to process spectral and spatial features simultaneously. A cross-scale feature fusion 10 block integrates information across multiple spatial resolutions. A classification head classifies the RGB images to output crop class labels.

Abstract: Video bandwidth optimization may be implemented using video resolution downscaling and upscaling in a video communications platform. Video bandwidth optimization may include receiving a first video stream comprising first multiple image frames at a first resolution. A video stream may be generated at a second resolution higher than the first resolution using a trained machine learning network. A user interface may display the video stream.

Abstract: This disclosure describes systems, methods, and devices related to deep learning-based video processing. A system may include a first neural network associated with generating kernel weights for the DLVP, the first neural network using a first hardware device; and a second neural network associated with filtering image pixels for the DLVP, the second neural network using a second hardware device, wherein the first neural network receives image data and generates the kernel weights based on the image data, and wherein the second neural network receives the image data and the kernel weights, and generates filtered image data based on the image data and the kernel weights.

Abstract: Apparatuses, systems, and techniques are presented to train one or more neural networks. In at least one embodiment, one or more neural networks are trained based, at least in part, on two or more versions of an image, wherein each of the two or more versions of the image are to be synthetically generated independently.

Abstract: Denoising of magnetic resonance (MR) images can be achieved using a deep neural network and an image acquisition process that uses multi-NEX (Number of Excitations) or multi-NSA (Number of Signal Averages or Acquisitions) to produce two or more complex-valued images of a region of interest. The set of images resulting from the image acquisition process can be input to a deep neural network that has been trained to produce a denoised MR image from a set of multi-NEX or multi-NSA images. The deep neural network can be implemented using a two-dimensional or three-dimensional convolutional neural network to match the dimensionality of the input images. Training of the denoising neural network can use real MR images.

Abstract: A neural network is trained and implemented to simultaneously remove noise and artifacts from medical images using a Generalized noise and Artifact Reduction Network (“GARNET”) method for training a convolutional neural network (“CNN”) or other suitable neural network or machine learning algorithm. Noise and artifact realizations from phantom images are used to synthetically corrupt images for training. Corrupted and uncorrupted image pairs are used for training GARNET. Following the training phase, GARNET can be used to improve image quality of routine medical images by way of noise and artifact reduction.

Abstract: A method and system for image processing are disclosed. A method for image processing by one or more central processing units (CPU) may include detecting an overlap pattern for a set of slice images of a porous media sample, based on the overlap pattern, determining a set of overlap distances for the set of slice images of the porous media sample, and registering, based on at least the set of overlap distances, a composite image comprising any of the set of slice images of the porous media sample.

Abstract: A biometric imaging device is provided with: placement unit that supports a biological object, imaging unit that takes images of the biological object, and determining unit that determines whether or not the biological object is placed on the placement unit, based on whether or not a feature of the biological object has changed at a place where the placement unit and the biological object are in contact, in the image(s).

Abstract: The present invention relates to an intelligent recognition method for job contents of barbershop employees, which belongs to the technical field of machine vision, comprising the following steps: S1: setting installation locations and installation conditions for collection devices; S2: establishing a label set of barbershop employee faces and customer faces, and training a face recognition model; S3: establishing an action label set related to items, tools and people, and training a barbershop action behavior recognition model; S4: conducting action behavior recognition for a practical barbershop service scenario by using the trained face recognition model and the barbershop action behavior recognition model; and constructing behavior sequences of “action pairs” involving elements from customers, employees and actions; S5: establishing a job content recognition label; and constructing a deep neural network model for job content recognition to determine service job contents of barbershop employees for custome

Abstract: An infrared vision positioning method for concrete pumping liquid level in tube of steel tubular arch bridge is provided. In this method, the perspective geometry algorithm is used to construct the pose transformation matrix to obtain the relative pose between the camera and the arch bridge steel pipe, the image is transformed into orthography by perspective transformation, and the real coordinates of the liquid level in the infrared image can be directly solved by the scale, so as to realize the accurate positioning of the liquid level; the proposed method is simple in structure and can realize real-time operation, which greatly improves the calculation efficiency of coordinates in infrared images, compared with the conventional knocking method, the advantage is that it does not require staff in an overhead operating, and the feedback of the liquid level position can be faster and more efficient.

Abstract: Methods and systems for identifying areas of interest in an image are disclosed. To manage identification of areas of interest in an image, a trained inference model may generate inferences based on the pixels of the image. The inferences may include areas of interest that contributed to the generation of the inferences. Some areas of interest may be highly relevant to the inferences and may be classified as primary areas of interest. Auxiliary areas of interest may also be identified using a trained inference model. Auxiliary areas of interest may be obtained by calculating gradients for each pixel that contributed to the identification of the primary areas of interest. By rank ordering the pixels, pixels with the highest contribution to the identification of the primary areas of interest may be identified. Proximate groupings of these pixels may be classified as auxiliary areas of interest in the image.

Abstract: A system and method for image correction is provided. The method includes: receiving an original image including a metal artifact; pre-correcting the original image; generating correction data based on the original image and the pre-corrected image; weighting the original image and the pre-corrected image based on the correction data and a regularity degree of the metal artifact, the regularity degree of the metal artifact being associated with actual projection data of the metal artifact and ideal projection data of the metal artifact; and generating a corrected image based on the weighting.

Abstract: The present disclosure is directed, among other things, to automated systems and methods for analyzing, storing, and/or retrieving information associated with biological objects having irregular shapes. In some embodiments, the systems and methods partition an input image into a plurality of sub-regions based on localized colors, textures, and/or intensities in the input image, wherein each sub-region represents biologically meaningful data.

Abstract: Provided is an information processing method that includes: performing a conversion process on the first feature information to obtain a first conversion result; performing the conversion process on the second feature information to obtain a second conversion result; performing a projection process on the first conversion result to obtain a first projection result; performing the projection process on the second conversion result to obtain a second projection result; calculating an error between the first projection result and the second projection result; and training the second model to reduce the error. The conversion process produces an error between the first projection result and the second projection result that is greater than the error between a projection result obtained by performing the projection process on the first feature information and a projection result obtained by performing the projection process on the second feature information.

Abstract: A first image frame in a first resolution format comprising image signal intensity values mapped to first pixel locations has sampling points offset from the centers of first pixel locations according to associated jitter vectors. The image signal intensity values are mapped to second pixel locations in a second image frame in a second resolution format based at least in part on the jitter vectors, the second resolution format being higher resolution than the first resolution format. The mapped image signal intensity values are combined with image signal intensity values of an accumulated history image buffer according to coefficients predicted by a neural network such as based on magnitudes of the jitter vectors. Interpolated pixel image intensities are added to the accumulated history buffer for empty or null pixel locations in the second image frame in the second image format. Upsampling artifacts such as checkerboarding and aliasing are reduced.

Abstract: A method for sparse optical flow based tracking in a computer vision system is provided that includes detecting feature points in a frame captured by a monocular camera in the computer vision system to generate a plurality of detected feature points, generating a binary image indicating locations of the detected feature points with a bit value of one, wherein all other locations in the binary image have a bit value of zero, generating another binary image indicating neighborhoods of currently tracked points, wherein locations of the neighborhoods in the binary image have a bit value of zero and all other locations in the binary image have a bit value of one, and performing a binary AND of the two binary images to generate another binary image, wherein locations in the binary image having a bit value of one indicate new feature points detected in the frame.

Abstract: A method includes acquiring an MRI complex signal having a plurality of complex echoes during an SWI sequence. The method includes phase filtering each complex echo of the plurality of complex echoes. The method also includes generating a respective phase image and a respective magnitude image from each phase filtered complex echo. The method further includes combining separately the respective magnitude images of the plurality of complex echoes with each other to generate a combined magnitude image and the respective phase images of the plurality of complex echoes with each other to generate a combined phase image. The method includes generating a complex image from both the combined magnitude image and the combined phase image. The method includes utilizing a deep learning-based denoising network to denoise the complex image to generate a denoised complex image.

Abstract: A control device 20 includes an image acquisition unit 203 configured to acquire a radiographic image obtained by irradiating a subject F with radiation and capturing an image of the radiation passing through the subject F, a noise map generation unit 204 configured to derive an evaluation value obtained by evaluating spread of a noise value from a pixel value of each pixel in the radiographic image on the basis of relationship data indicating a relationship between the pixel value and the evaluation value and generate a noise map that is data in which the derived evaluation value is associated with each pixel in the radiographic image, and a processing unit 205 configured to input the radiographic image and the noise map to a trained model 207 constructed in advance through machine learning and execute image processing of removing noise from the radiographic image.

Abstract: This invention provides a system and a computer implemented method to match gemstones with records in a gemstone database. In one embodiment, said computer implemented method comprises the steps of: a) Acquiring a source image of a gemstone; b) Receiving an input for identifying a corresponding gemstone in said gemstone database; c) Retrieving a set of reference data; d) Extracting gemstone features from said source image; e) Performing alignment of said gemstone features with said set of reference data and performing a homography transformation to map said gemstone features onto a standard plane to obtain a transformed source image from said source gemstone; f) Obtaining a transformed reference image from said set of reference data; g) Computing a similarity value by matching said transformed source image against said transformed reference image; and h) Determining whether said gemstone in question matches said corresponding gemstone based on said similarity value.

Abstract: A computer-implemented method is provided for processing images. The method can include down-sampling a plurality of first images having a first resolution for obtaining a plurality of second images having a second resolution and training an artificial neural network model to process an input image and output an output image having a higher resolution than the input image.