Abstract: A resource transfer method includes obtaining, by using a target application, a resource transfer request triggered by a target user, where the resource transfer request includes verification information used to perform resource transfer processing and identity feature information of the target user. By using the target application, verifying the verification information, and invoking, if the verification succeeds by using the target application, a local device management rule, and determining, by using the local device management rule, whether the identity feature information of the target user matches identity feature information of a pre-registered user. If the identity feature information of the target user matches the identity feature information of the pre-registered user, sending the resource transfer request to a first server corresponding to the target application to trigger, based on the resource transfer request, the first server to perform resource transfer processing.

Abstract: An example apparatus for detecting objects in video frames includes a receiver to receive a plurality of video frames from a video camera. The apparatus also includes a first still image object detector to receive a first frame of the plurality of video frames and calculate localization information and confidence information for each potential object patch in the first frame. The apparatus further includes a second still image object detector to receive an adjacent frame of the plurality of video frames adjacent to the first frame and calculate localization information and confidence information for each potential object patch in the adjacent frame. The apparatus includes a similarity detector trained to detect paired patches between the first frame and the adjacent frame based on a comparison of the detected potential object patches.

Abstract: A method, decoder, and apparatus are provided. Responsive to a current block being a MIP predicted block, it is determined whether it has one or multiple transform blocks. A MIP weight matrix to be used to decode the current block is determined based on a MIP prediction mode. Responsive to the MIP predicted block having one transform block, the MIP predicted block is derived based on the MIP weight matrix and previously decoded elements in the bitstream. Responsive to the MIP predicted block having multiple transform blocks: deriving a first MIP predicted block is derived based on the MIP weight matrix and previously decoded elements in the bitstream and remaining MIP predicted blocks are derived based further on decoded elements in at least one decoded transform block of the current block. The MIP predicted block(s) are output for subsequent processing.

Abstract: Provided are a method and apparatus for evaluating a degree of frequency regulation urgency of a generator set, a power system and a storage medium.

Abstract: A double-sided loudspeaker includes a first speaker base and a second speaker base assembled with each other and provided with a ventilation hole; a first side magnetic plate injection molded at the first speaker base or the second speaker base; a second side magnetic plate injection molded at the first speaker base or the second speaker base; a magnetic member disposed between the first side magnetic plate and the second side magnetic plate and including a main magnetic plate, a first magnetic unit and a second magnetic unit; a first diaphragm disposed at the first speaker base; a first voice coil disposed at the first diaphragm and movably arranged at the first magnetic unit; a second diaphragm disposed at the second speaker base; and a second voice coil disposed at the second diaphragm and movably arranged at the second magnetic unit.

Abstract: Systems and methods process images to determine a skin condition severity analysis and to visualize a skin analysis such as using a deep neural network (e.g. a convolutional neural network) where a problem was formulated as a regression task with integer-only labels. Auxiliary classification tasks (for example, comprising gender and ethnicity predictions) are introduced to improve performance. Scoring and other image processing techniques may be used (e.g. in assoc. with the model) to visualize results such as highlighting the analyzed image. It is demonstrated that the visualization of results, which highlight skin condition affected areas, can also provide perspicuous explanations for the model. A plurality (k) of data augmentations may be made to a source image to yield k augmented images for processing. Activation masks (e.g. heatmaps) produced from processing the k augmented images are used to define a final map to visualize the skin analysis.

Abstract: An optical soft skin system for multimodal sensing comprising: a flexible waveguide substrate (10) being able to change its light wave propagation when encountering different deformation patterns; at least one light source (12) located within the waveguide substrate (10) arranged so as to transmit light into the waveguide substrate (10); at least one light receptor (14) located within the waveguide substrate (10) to receive light that has passed through the waveguide substrate (10) and to provide a dataset of the properties of the received light; and a processor (20) for decoding the dataset into at least one of strain and displacement through a computational model to determine a sensing modality of the waveguide substrate (10).

Abstract: A data storage system with intelligent power management includes a plurality of data storage devices and a controller. Each data storage device is capable of operating in one of (N+1) power saving functions where N is an integer larger than 1. The (N+1) power saving functions sequentially correspond to from the 0th to the Nth power saving levels. The controller reads a user-setting power saving level (I) where I is an integer index ranging from 0 to N. The controller reads a current power saving level (J) of a current power saving function of one of the plurality of data storage devices where J is an integer index ranging from 0 to N. The controller controls said one data storage device to operate in one power saving function among the (N+1) power saving functions according to the user-setting power saving level (I) and the current power saving level (J).

Abstract: The present disclosure provides a log information obtaining system and method. The log information obtaining system includes a baseboard management controller (BMC), a storage device, a first interface, a controller and a first selector. The storage device stores log information of the electronic device. The controller obtains a first signal of the BMC and determines the working state of the BMC, the controller outputs a first selection signal when the working state of the BMC is abnormal. The first selector selects the first output terminal to output the log information received from the first input terminal to the external device. Even if the BMC cannot work normally, it can still obtain the log information of the electronic device and debug the electronic device.

Abstract: A data center security control module able to connect with motherboards of different platforms such as Intel platform, AMD platform, and Ampere platform includes a baseboard management controller (BMC), a chip selection module, and a control module. The BMC outputs an alarm signal. The chip selection module includes N chip selection units, the chip selection unit outputting alarm signal to an external motherboard which is connected to the data center security control module. The control module obtains information as to motherboard type being Intel platform, or AMD platform, or Ampere platform and outputs a control signal to the chip selection module according to the motherboard type to start the chip selection unit associated with such motherboard type. The application also provides a data center security control module control method.

Abstract: A board mis-insertion prevention circuit includes a first detection circuit, a second detection circuit, and a control module. The first detection circuit detects whether a board is in a first state and outputs a first detection signal when the board is in the first state. The second detection circuit detects whether the board is in a second state when the board is not in the first state, and outputs a second detection signal when the board is in the second state. The control module receives the first detection signal or the second detection signal and can allow or disconnect a power supply to the board according to the first detection signal or the second detection signal.

Abstract: The present disclosure provides a method for treating arthritis by using a stem cell preparation. The stem cell preparation of the present disclosure can effectively delay cartilage degeneration caused by arthritis, and it is confirmed by whole blood analysis and blood biochemical analysis that the stem cell preparation in the form of three-dimensional stem cell spheres provides a safe treatment for arthritis. The present disclosure also provides a method for preparing the stem cell preparation.

Abstract: A method detects mechanical equipment parts. The method includes: obtaining an image of a part; extracting a feature from the image using a machine learning model, identifying a type of surface defect on the basis of the feature to obtain an identification result; and determining whether to replace the part on the basis of the identification result and a predetermined standard of the part. The method reduces the difficulty of detecting a part, can accurately identify a surface defect of the part and determine whether the part needs to be replaced, thereby improving the work efficiency, and shortens the time for mechanical equipment to stop operating for maintenance, thus improving the operating efficiency of the mechanical equipment. The method is automatically executed by a computer, thereby avoiding manually checking errors, improving the accuracy of detection results, and thus improving the reliability of operation of the mechanical equipment.

Abstract: With Convolutional Neural Networks (CNN), facial alignment networks (FAN) have achieved significant accuracy on a wide range of public datasets, which comes along with larger model size and expensive computation costs, making it infeasible to adapt them to real-time applications on edge devices. There is provided a model compression approach on FAN using One-Shot Neural Architecture Search to overcome this problem while preserving performance criteria. Methods and devices provide efficient training and searching (on a single GPU), and resultant models can deploy to run real-time in browser-based applications on edge devices including tablets and smartphones. The compressed models provide comparable cutting-edge accuracy, while having a 30 times smaller model size and can run 40.7 ms per frame in a popular browser on a popular smartphone and OS.

Abstract: A lutein-containing ophthalmic composition, wherein the ophthalmic composition is mainly consisting of a natural anti-inflammatory substance, a substance with increasing liquid viscosity, and an artificial tear. The substance with increasing liquid viscosity could apply the ophthalmic composition to elongate the retention time on the ocular surface, thereby increasing the retention time of anti-inflammatory substance on the ocular surface for treating dry eye syndrome and inhibiting ocular surface inflammation.

Abstract: There are provided systems and methods for facial landmark detection using a convolutional neural network (CNN). The CNN comprises a first stage and a second stage where the first stage produces initial heat maps for the landmarks and initial respective locations for the landmarks. The second stage processes the heat maps and performs Region of Interest-based pooling while preserving feature alignment to produce cropped features. Finally, the second stage predicts from the cropped features a respective refinement location offset to each respective initial location. Combining each respective initial location with its respective refinement location offset provides a respective final coordinate (x,y) for each respective landmark in the image. Two-stage localization design helps to achieve fine-level alignment while remaining computationally efficient.

Abstract: There are provided systems and methods for facial landmark detection using a convolutional neural network (CNN). The CNN comprises a first stage and a second stage where the first stage produces initial heat maps for the landmarks and initial respective locations for the landmarks. The second stage processes the heat maps and performs Region of Interest-based pooling while preserving feature alignment to produce cropped features. Finally, the second stage predicts from the cropped features a respective refinement location offset to each respective initial location. Combining each respective initial location with its respective refinement location offset provides a respective final coordinate (x,y) for each respective landmark in the image. Two-stage localization design helps to achieve fine-level alignment while remaining computationally efficient.

Abstract: There is described a deep learning supervised regression based model including methods and systems for facial attribute prediction and use thereof. An example of use is an augmented and/or virtual reality interface to provide a modified image responsive to facial attribute predictions determined from the image.