Abstract: A partner matching method in a costarring video is performed by a terminal. The terminal obtains a video recorded by a first user identifier and a video in which a second role that matches the first role is played and an associated second user identifier. After obtaining a total score of videos in which the second role is played by each second user identifier in each user type, the terminal ranks the videos in which the second role is played by the second user identifiers for each user type and displays a ranking result of the videos in which the second role is played for each user type. After obtaining a video selected from the ranking result, the terminal synthesizes a complete video from the selected video in which the second role is played and the video in which the first role is played.

Abstract: A method and apparatus for automated vertebra localization and identification in a 3D computed tomography (CT) volumes is disclosed. Initial vertebra locations in a 3D CT volume of a patient are predicted for a plurality of vertebrae corresponding to a plurality of vertebra labels using a trained deep image-to-image network (DI2IN). The initial vertebra locations for the plurality of vertebrae predicted using the DI2IN are refined using a trained recurrent neural network, resulting in an updated set of vertebra locations for the plurality of vertebrae corresponding to the plurality of vertebrae labels. Final vertebra locations in the 3D CT volume for the plurality of vertebrae corresponding to the plurality of vertebra labels are determined by refining the updated set of vertebra locations using a trained shape-basis deep neural network.

Abstract: A thermal protection switch device may include: a power circuit that includes a phase line and a neutral line; a current sensor electrically connected to the phase line to detect the current flow in the phase line; a temperature sensor thermally coupled to the phase line to detect the temperature of the phase line; and another temperature sensor thermally coupled to the neutral line to detect the temperature of the neutral line. The device also includes a trip circuit configured to interrupt the phase line when activated and a controller configured to receive output signals from the above current and temperature sensors. The controller may activate the trip circuit to interrupt the phase line in response to determining that certain conditions have occurred.

Abstract: A method for realizing precise gain control for a hybrid fibre amplifier, and a hybrid fibre amplifier, in which by an erbium-doped fibre amplifier firstly outputting a constant power, a comparable source signal optical power is provided for a raman fibre amplifier of a next stage. A feedback for the gain control may be formed by comparing a source signal optical power calculated after starting pumping of the Raman fibre amplifier and a source signal optical power detected after pumping stops, thereby greatly improving gain control precision of the Raman fibre amplifier. Moreover, the erbium-doped fibre amplifier parts of all the hybrid fibre amplifiers may simultaneously output a constant optical power, and the Raman amplifier parts of all the hybrid fibre amplifiers may simultaneously start calibration, so that the time for starting operation of the entire system may be improved greatly.

Abstract: The present embodiments relate to detecting multiple landmarks in medical images. By way of introduction, the present embodiments described below include apparatuses and methods for detecting landmarks using hierarchical feature learning with end-to-end training. Multiple neural networks are provided with convolutional layers for extracting features from medical images and with a convolutional layer for learning spatial relationships between the extracted features. Each neural network is trained to detect different landmarks using a different resolution of the medical images, and the convolutional layers of each neural network are trained together with end-to-end training to learn appearance and spatial configuration simultaneously. The trained neural networks detect multiple landmarks in a test image iteratively by detecting landmarks at different resolutions, using landmarks detected a lesser resolutions to detect additional landmarks at higher resolutions.

Abstract: A partner matching method in a costarring video is performed by a terminal. The terminal obtains a video recorded by a first user identifier and a video in which a second role that matches the first role is played and an associated second user identifier. After obtaining a total score of videos in which the second role is played by each second user identifier in each user type, the terminal ranks the videos in which the second role is played by the second user identifiers for each user type and displays a ranking result of the videos in which the second role is played for each user type. After obtaining a video selected from the ranking result, the terminal synthesizes a complete video from the selected video in which the second role is played and the video in which the first role is played.

Abstract: An audible ground fault detection device may include: a pulse-width modulation (PWM) circuit to produce a pulse signal in response to an occurrence of a ground fault; and a buzzer circuit to emit an audible sound in response to receipt of the pulse signal. The device may include a power circuit configured to transfer power from the power line to the PWM circuit when a ground fault occurs in a load of the power line, and not to transfer the power from the power line to the PWM circuit when no ground fault occurs in any load of the power line. When the PWM circuit has no power, it will not produce the pulse signal. The power circuit may also provide DC power to the buzzer circuit in response to the occurrence of the ground fault. The buzzer circuit may include a passive buzzer.

Abstract: This application discloses a multimedia file joining method performed by an apparatus. After obtaining a first video clip and a second video clip to be joined, the apparatus obtains an audio file corresponding to the first video clip and the second video clip. The audio file records the first start and end time points of the first video clip and the second start and end time points of the second video clip. The apparatus adjusts the first video clip to play the first video clip in a first time period indicated by the first and end time points, and adjusts the second video clip to play the second video clip in a second time period indicated by the second and end time points with the first time period not overlapping the second time period. Finally, the apparatus performs a joining operation on the adjusted first video clip and the adjusted second video clip, to obtain a joined video file.

Abstract: A data transmission acceleration method and related apparatuses are disclosed. A sanding node transmits a plurality of data packets to a receiving node at an initial transmission rate. Each data packet carries a random sequence number and a rolling sequence number. The random sequence number identifies a data part of the data packet, and the rolling sequence number indicates a transmission sequence of the data packet. The sending node receives a packet loss feedback from the receiving node. The packet loss feedback is generated after the receiving node detects a packet loss event according to a rolling sequence number of a received data packet. The sending node determines a random sequence number of a lost data packet based on the received packet loss feedback. The sending node retransmits a data packet corresponding to the determined random sequence number. The retransmitted data packet carries a new rolling sequence number.

Abstract: A method and apparatus for automated vertebra localization and identification in a 3D computed tomography (CT) volumes is disclosed. Initial vertebra locations in a 3D CT volume of a patient are predicted for a plurality of vertebrae corresponding to a plurality of vertebra labels using a trained deep image-to-image network (DI2IN). The initial vertebra locations for the plurality of vertebrae predicted using the DI2IN are refined using a trained recurrent neural network, resulting in an updated set of vertebra locations for the plurality of vertebrae corresponding to the plurality of vertebrae labels. Final vertebra locations in the 3D CT volume for the plurality of vertebrae corresponding to the plurality of vertebra labels are determined by refining the updated set of vertebra locations using a trained shape-basis deep neural network.

Abstract: Provided are a control method and system for a cascade hybrid amplifier, in which respective hybrid amplifiers in the cascade hybrid amplifier simultaneously start to implement a pump-starting process comprising: when the hybrid amplifier receives a request to start pumping, determining whether conditions are satisfied, if yes, determining stability of power of an input light of a Raman, starting pumping of an EDFA so that the EDFA enters into an APC operation mode; starting pumping of the Raman, and calculating a gain deviation according to the calculated input light powers before and after pump-starting of the Raman when no reflection alarm exists; and adjusting gain of the Raman according to the gain deviation, and switching to an AGC (automatic gain control) operation mode after the adjustment; and switching the EDFA to the AGC operation mode.

Abstract: A pixel having a variable exposure for a scene capture device, the pixel comprising a photodiode, a buffer and a memory element and a switching mechanism. The memory element is configured to store an exposure control bit and the switching mechanism configured to control a variable exposure period of the photodiode based on the exposure control bit and to reset a voltage on said photodiode to a reference voltage.

Abstract: In one embodiment, a method determines known features for existing face tracks that have identity labels and builds a database using these features. The face tracks may have multiple different views of a face. Multiple features from the multiple faces may be taken to build the face models. For an unlabeled face track without identity information, the method determines its sampled features and finds labeled nearest neighbor features with respect to multiple feature spaces from the face models. For each face in the unlabeled face track, the method decomposes the face as a linear combination of its neighbors from the known features from the face models. Then, the method determines weights for the known features to weight the coefficients of the known features. Particular embodiments use a non-linear weighting function to learn the weights that provides more accurate labels.

Abstract: In one embodiment, a method determines a video including ad slots inserted within the video. The method generates a caption curve for a caption file of caption segments for a video based on start and stop times for caption segments in the caption file. The caption segments in the caption file were generated for the video without including ad slots. Then, the method determines a speech velocity for the video using the caption file and revises the caption curve based on the speech velocity and a number of characters in caption segments in the caption file. A speech probability curve is determined based on audio of the video and the method correlates the speech probability curve to the revised caption curve to align the caption segments of the caption file with speech of the video.

Abstract: The present embodiments relate to detecting multiple landmarks in medical images. By way of introduction, the present embodiments described below include apparatuses and methods for detecting landmarks using hierarchical feature learning with end-to-end training. Multiple neural networks are provided with convolutional layers for extracting features from medical images and with a convolutional layer for learning spatial relationships between the extracted features. Each neural network is trained to detect different landmarks using a different resolution of the medical images, and the convolutional layers of each neural network are trained together with end-to-end training to learn appearance and spatial configuration simultaneously. The trained neural networks detect multiple landmarks in a test image iteratively by detecting landmarks at different resolutions, using landmarks detected a lesser resolutions to detect additional landmarks at higher resolutions.

Abstract: In one embodiment, a method selects a set of accounts related to a video using a first classifier that is trained to identify the set of accounts from a plurality of accounts and selects a token used to identify the video. A second classifier is trained to identify short messages that are related to the video using a first set of short messages from the accounts related to the video and the second set of short messages including the tokens. A third set of short messages is identified that is related to the video from a plurality of short messages. The method normalizes words in the set of short messages using a mixed model including a lexicon model and a translation model. Then, the method generates a set of ratings for the set of short messages and uses the set of ratings in a service for a video delivery service.