Abstract: This application relates to a machine learning model training method and apparatus, and an expression image classification method and apparatus. The machine learning model training method includes: obtaining a machine learning model that includes a model parameter and that is obtained through training according to a general-purpose image training set; determining a sample of a special-purpose image and a corresponding classification label; inputting the sample of the special-purpose image to the machine learning model, to obtain an intermediate classification result; and adjusting the model parameter of the machine learning model according to a difference between the intermediate classification result and the classification label, continuing training, and ending the training in a case that a training stop condition is met. The solutions provided in this application improve the training efficiency of the machine learning model.

Abstract: An automatic image annotation system receives a reference image with one or more parts annotated along with one or more query images and automatically identifies portions from the query images that are similar to the annotated parts of the reference image. The S-matrices of the reference image and the query images are obtained via singular value decomposition (SVD). Lower-dimensional images are also obtained for the reference image and the query images using a pre-trained deep learning model. The S-matrices and the lower-dimensional images of the corresponding images are combined to generate vector representations. A distance metric is calculated for the vector representation of the reference image with that of the query image. A preliminary output image with a preliminary annotation is initially generated. The preliminary annotation is further optimized to generate an optimized annotation that adequately covers the region of interest (ROI).

Abstract: A system and method for generating a bounding box for an object in proximity to a vehicle are disclosed. The method includes: receiving a three-dimensional (3D) point cloud representative of an environment; receiving a two-dimensional (2D) image of the environment; processing the 3D point cloud to identify an object cluster of 3D data points for a 3D object in the 3D point cloud; processing the 2D image to detect a 2D object in the 2D image and generate information regarding the 2D object from the 2D image; and when the 3D object and the 2D object correspond to the same object in the environment: generating a bird's eye view (BEV) bounding box for the object based on the object cluster of 3D data points and the information from the 2D image.

Abstract: This application discloses an image feature recognition method performed at a computing device. The computing device obtains a first neural network model by training parameters in a second neural network model using a first training set and a second training set consecutively, image features of training pictures in the first training set being marked and image features of training pictures in the second training set being not marked. After obtaining a first neural network model, the computing device obtains a recognition request for recognizing an image feature in a target picture. The computing device then recognizes the image feature in the target picture by applying the target picture to the first neural network model. Finally the computing device returns a first recognition result of the first neural network model, the first recognition result indicating an image feature (for example, a pathological feature) recognized in the target picture.

Abstract: Low power consumption baseline tracking systems and methods for automatically tracking a baseline input into a capacitive sensor having a plurality of transmitter electrodes and a plurality of receiver electrodes. A partial baseline scan is captured by driving all or a portion of the plurality of transmitter electrodes simultaneously and detecting receiver signals from a subset of the at least one receiver electrode simultaneously, and the partial baseline scan is compared with a stored baseline image. When a difference between the captured partial baseline scan and the stored baseline image exceeds a threshold value, a full baseline image scan is acquired, and the stored baseline image is updated.

Abstract: Embodiments of the present disclosure disclose a method and apparatus for generating position information, a device and a medium. A specific embodiment of the method includes: acquiring an image and vehicle position information, wherein the image includes a target element; inputting the image into a pre-established depth map generation model to obtain a first depth map, wherein the focal length of sample images of sample data used during the training of the model is a sample focal length; generating a second depth map based on the sample focal length, the first depth map, and an estimated focal length of the image; determining depth information of the target element according to element position information of the target element in the image and the second depth map; and generating position information of the target element based on the vehicle position information and the depth information of the target element.

Abstract: The present disclosure relates to systems, methods, and computer-readable media for selectively identifying pixel data to provide as an input to an image processing model based on motion data associated with the content of a digital video. For example, systems disclosed herein include receiving a compressed digital video and decompressing the compressed digital video to generate a decompressed digital video. The systems disclosed herein further include extracting or otherwise identifying motion data while decompressing the compressed digital video. The systems disclosed herein also include analyzing the motion data to determine a subset of pixel data from the decompressed digital video to provide as input to an image processing model trained to generate an output based on input pixel data.

Abstract: According to an embodiment of the present invention, there is provided a method of screening a prostate cancer patient by optical image analysis of a circulating tumor cell marker and a prostate-specific membrane antigen.

Abstract: A method for automatic glue-spraying of stringers and inspection of glue-spraying quality based on measured data. Three-dimensional (3D) point cloud data of a stringer-skin assembly is collected by 3D laser scanner, and then processed by denoising and sampling. Feature points of an intersection line of a site to be glued of the stringer-skin assembly are extracted by K-means clustering method based on Gaussian mapping, and a minimum spanning tree is constructed based on a set of the extracted feature points. A connected region is established to obtain an initial feature intersection line of the string-skin assembly, which is optimized by random sample consensus algorithm to remove redundant small branch structures to obtain the actual glue-spraying trajectory. The quality of the glue sprayed on the stringer-skin assembly is inspected by line laser to determine positions of the defects, which are then subjected to secondary glue-spraying.

Abstract: A medical imaging apparatus (1) and method for imaging a light-sensitive object (2), such as tissue (3) of the eye (4), displays the object (2) in spectral bands (30) of the visible light range (37) in which the object is sensitive and cannot be illuminated. The apparatus (1) has a camera (8) for capturing input images (10) in which a spectral signature (36) of the object is represented by an input set (15) of spectral bands (16). The apparatus generates output images (22) in which the spectral signature (60) in the visible light range (37) is represented by an output set (29) of spectral bands (30). Replacing at least one input-only spectral band (38) in the input set (15) by at least one output-only spectral band (39) in the output set (29) enables capture of input images (10) in spectral bands (16) where the object (2) has less sensitivity, such as in the non-visible light-range (18), and display of the object (2) in spectral bands (30) where the object would be too sensitive for illumination (40).

Abstract: Provided is an oil leakage detection apparatus by which oil leakage detection is performed even for colorless oil with high detection accuracy and without complicating the apparatus. The oil leakage detection apparatus of the invention includes a distance measurement unit configured to measure a distance to the oil-input machine, an ultraviolet light source configured to irradiate the oil-input machine with ultraviolet light, a color imaging unit configured to capture an image of the oil-input machine irradiated with ultraviolet light, an image processing unit configured to diagnose oil leakage of the oil-input machine based on the distance measured by the distance measurement unit and the captured image of the color imaging unit, and a display unit configured to display a processed image processed by the image processing unit.

Abstract: The present application discloses a method and a device for analyzing the water content of skin by means of a skin image. The method comprises performing skin feature analysis on an acquired skin image, and obtaining the water content of skin on the basis of the skin features, wherein the skin features comprise a glossiness level and a smoothness level. The present invention clearly shows a reduction in implementation costs when compared to the prior art. The invention also achieves rapid testing.

Abstract: The invention relates to a handheld device and method for determining a status of a plant. The device includes a multi pixel digital colour sensor, a light source arranged for providing broadband illumination, wherein the light source and the multi pixel digital colour sensor are arranged in substantially the same plane, a light guide for guiding the light from said light source into the direction of the multi pixel digital colour sensor, a sample space, provided between the multi pixel digital colour sensor and the light source, for insertion of at least a part of the plant therein, and a processing unit configured for controlling at least the multi pixel digital colour sensor and the light source.

Abstract: Some embodiments include a method, comprising: receiving an image representing a branching structure; determining a starting feature of the branching structure; selecting a subregion of the image based on the starting feature; segmenting the branching structure in the subregion; generating a set of next features based on the segmented branching structure; and for each of the next features, repeating the selecting of the subregion based on the next feature, the segmenting of the branching structure, and the generating of the set of next features.

Abstract: A PET image reconstruction method, including: 1) injecting a PET radioactive tracer into a biological tissue, scanning by a PET device, and detecting and counting coincidence photons to obtain an original protection data matrix; 2) establishing a measurement equation model; 3) splitting the reconstruction problem into a first sub-problem and a second sub-problem; 4) solving the first sub-problem by a filtered back-projection layer, solving the second sub-problem by an improved denoising convolutional neural network, where the filtered back-projection layer and the improved denoising convolutional neural network are connected in series to form a filtered back-projection network (FBP-Net); 5) inputting original projection data into the FBP-Net, and using an image as a tag to adjust parameters of the FBP-Net to reduce an error between an output of the FBP-Net and the tag; and 6) inputting projection data to be reconstructed into the trained FBP-Net to obtain a desired reconstructed image.

Abstract: An image processing apparatus includes search area calculation means for calculating a search area size on the basis of a pixel size of at least one image among a first image and a second image, position obtaining means for obtaining a position of interest on the first image and a correspondence position on the second image corresponding to the position of interest, search area setting means for setting a search area made up of the search area size in a surrounding of the correspondence position on the second image, and difference means for deciding a difference value corresponding to the position of interest on a basis of a density value of the position of interest on the first image and density values of a plurality of positions in the search area on the second image.

Abstract: An image processing apparatus includes an image corrector and an output circuit. The image corrector inputs image data. The image corrector processes the image data input by the image corrector. The image corrector arranges an input image indicated by the image data input by the image corrector on a first plane. The image corrector maps each pixel included in a part or whole of a region of the input image arranged on the first plane onto a curved surface. The image corrector maps each point mapped on the curved surface onto a second plane so as to generate new image data. The output circuit outputs the image data processed by the image corrector.

Abstract: The present disclosure provides an image recognition method and apparatus, a device and a non-volatile computer storage medium. In embodiments of the present disclosure, it is feasible to obtain the to-be-recognized image of a designated size, extract the different-area image from the to-be-recognized image, and obtain the image feature of the different-area image according to the different-area image, so as to obtain the recognition result of the to-be-recognized image, according to the image feature of the different-area image and a preset template feature. In this way, recognition processing can be performed for images in a limited number of classes without employing the deep learning method based on hundreds of thousands of even millions of training samples.

Abstract: Vehicles and systems and methods for determining a viewing direction of a user are disclosed. In one embodiment, a system for detecting a viewing direction of a user includes a marker arranged at a fixed location, and an optical assembly. The optical assembly includes a first optical sensor arranged to view the marker, and a second optical sensor arranged to view the user. The first optical sensor detects the marker for determining a spatial location of the optical assembly with respect to the marker, and the second optical sensor detects an eye of the user for determining an eye angle of the user.

Abstract: Three-dimensional (3D) imaging systems and methods are described for implementing virtual grading of package walls in commercial trailer loading. The 3D imaging systems and methods comprise capturing, by a 3D-depth camera, 3D image data of a vehicle storage area. A package wall is determined, based on the set of 3D image data and by a 3D data analytics application (app) executing on one or more processors communicatively coupled to the 3D-depth camera, a package wall within the vehicle storage area. A wall grade is assigned to the package wall, where the package wall is defined within the set of 3D image data by a plurality of packages each having a similar depth dimension, and wherein the package wall has a wall height.