In one embodiment, a method includes determining characteristics of one or more areas in an image by analyzing pixels in the image, computing a sampling density for each of the one or more areas in the image based on the characteristics of the one or more areas, generating samples corresponding to the image by sampling pixels in each of the one or more areas according to the associated sampling density, and providing the samples to a machine-learning model as an input, where the machine-learning model is configured to reconstruct the image by processing the samples.

Handwriting verification methods and related computer systems, and handwriting-based user authentication methods and related computer systems are disclosed. A handwriting verification method comprises obtaining a handwriting test sample containing a plurality of available parameters, extracting geometric parameters, deriving geometric features comprising an x-position value and a y-position value for each of a plurality of feature points in the test sample, performing feature matching between geometric features of the test sample and a reference sample, determining a handwriting verification result based at least in part on the feature matching, and outputting the handwriting verification result. The geometric features may further comprise values derived from the geometric parameters, such as direction and curvature values. The handwriting verification result can be further based on a count of unlinked feature points. Handwriting-based user authentication methods can employ such handwriting verification methods, or other handwriting verification methods.

An image processing method includes: obtaining a first target face image in a reference expression state; obtaining user face key point offset information that includes offsets between face key points of a user in a target expression state and face key points of the user in the reference expression state; adjusting face key points of the first target face image in the reference expression state according to the user face key point offset information; and attaching, in a to-be-processed image including a second target face image, an expression texture image of the adjusted first target face image to the second target face image, the second target face image and the first target face image belonging to a same target.

A method for detecting defects in working CNC tools in real time, implemented in an electronic device, includes acquiring sounds of operation of a tool during a cutting or other operation process and dividing the acquired cutting sounds into a plurality of recordings of audio according to a preset time interval. Time-frequency features of the plurality of recordings of audio are acquired according to multiple feature transformation methods and a fusion feature image of the cutting sound is formed according to the extracted time-frequency features. A tool detection model is generated by training the fusion feature image, and any defects of the tool and any defect types the tool has are detected according to the tool detection model.

A system and method are provided. The system includes a processor. The processor is configured to generate a response map for an image, using a four stage convolutional structure. The processor is further configured to generate a plurality of landmark points for the image based on the response map, using a shape basis neural network. The processor is additionally configured to generate an optimal shape for the image based on the plurality of landmark points for the image and the response map, using a point deformation neural network. A recognition system configured to identify the image based on the generated optimal shape to generate a recognition result of the image. The processor is also configured to operate a hardware-based machine based on the recognition result.

Two system-level bronchoscopy guidance solutions are presented. The first incorporates a global-registration algorithm to provide the physician with updated navigational and guidance information during bronchoscopy. The system can handle general navigation to a region of interest (ROI), as well as adverse events, and it requires minimal commands so that it can be directly controlled by the physician. The second solution visualizes the global picture of all the bifurcations and their relative orientations in advance and suggests the maneuvers needed by the bronchoscope to approach the ROI. Guided bronchoscopy results using human airway-tree phantoms demonstrate the potential of the two solutions.

The present disclosure provides a model training method, an identification method, device, storage medium and program product, relating to computer vision technology and deep learning technology. In the solution provided by the present application, the image is deformed by the means of deforming the first training image without label itself, and the first unsupervised identification result is obtained by using the first model to identify the image before deformation, and the second unsupervised identification result is obtained by using the second model to identify the image after deformation, and the first unsupervised identification result of the first model is deformed, thus a consistency loss function can be constructed according to the second unsupervised identification result and the scrambled identification result. In this way, it is able to enhance the constraint effect of the consistency loss function and avoid destroying the scene semantic information of the images used for training.

A fingerprint authentication device includes: a sensor unit configured to output a sensing signal by sensing a fingerprint; an image processing unit configured to generate a fingerprint image based on the sensing signal; a storage unit configured to store a template including an enrolled image; and a learning unit configured to generate a first pseudo image and add the first pseudo image to the template.

Treatment trajectory guidance systems and methods are provided. In one embodiment, the method for treatment trajectory guidance in a patient's brain includes obtaining a three- dimensional (3D) brain model that includes a model of an anatomy, the model of the anatomy including a plurality of feature points; modifying the 3D brain model based on magnetic resonance (MR) data of the patient's brain from a magnetic resonance imaging (MRI) device to obtain a plurality of modified feature points on a modified model of the patient's anatomy in the patient's brain; displaying on a display a first planned trajectory for treating the patient's anatomy based on the plurality of modified feature points; and displaying, on the display, a first estimated treatment result for the first planned trajectory.

Systems and methods of improving alignment in dense prediction neural networks are disclosed. A method includes identifying, at a computing system, an input data set and a label data set with one or more first parts of the input data set corresponding to a label. The computing system processes the input data set using a neural network to generate a predicted label data set that identifies one or more second parts of the input data set predicted to correspond to the label. The computing system determines an alignment result using the predicted label data set and the label data set and a transformation of the one or more first parts, including a shift, rotation, scaling, and/or deformation, based on the alignment result. The computing system computes a loss score using the transformation, label data and the predicted label data set and updates the neural network based on the loss score.