Abstract: A method of performing automatic tuning on a deep learning model includes: utilizing an instruction-based learned cost model to estimate a first type of operational performance metrics based on a tuned configuration of layer fusion and tensor tiling; utilizing statistical data gathered during a compilation process of the deep learning model to determine a second type of operational performance metrics based on the tuned configuration of layer fusion and tensor tiling; performing an auto-tuning process to obtain a plurality of optimal configurations based on the first type of operational performance metrics and the second type of operational performance metrics; and configure the deep learning model according to one of the plurality of optimal configurations.

Abstract: An online reconfigurable battery system with current surge protection modules (SPM), including: a plurality of battery module strings connected in parallel. The battery module string further includes: an SPM and a plurality of enable/bypass battery modules (EBM) connected in series; where the EBM further includes: a battery, a first switch, and a second switch; the first switch and the battery are connected in series, and then connected to the second switch in parallel to form an EBM that can be enabled or bypassed; the SPM further includes: a variable resistor, a third switch, and a fourth switch; the third switch and the variable resistors are connected in series, and then connected in parallel with the fourth switch to buffer the surge current.

Abstract: An online reconfigurable battery system with current surge protection modules (SPM), including: a plurality of battery module strings connected in parallel. The battery module string further includes: an SPM and a plurality of enable/bypass battery modules (EBM) connected in series; where the EBM further includes: a battery, a first switch, and a second switch; the first switch and the battery are connected in series, and then connected to the second switch in parallel to form an EBM that can be enabled or bypassed; the SPM further includes: a variable resistor, a third switch, and a fourth switch; the third switch and the variable resistors are connected in series, and then connected in parallel with the fourth switch to buffer the surge current.

Abstract: A power supply system with UPS, PCS and circuit diagnosis capabilities is disclosed, including: a DC-bus connected to a voltage/current (V/I) meter, a battery energy storage system (ESS) container, a power conditioning system (PCS), at least one current translation unit, and an energy management controller (EMC), wherein, the V/I meter is used to monitor the voltage and current of the DC-bus, and the PCS performs bi-directional conversion between the DC current from the DC-bus and the AC power from an external distribution panel, the current translation unit translate the DC current from the DC-bus into a voltage suitable for at least a critical load, and the EMC controls the operation of the V/I meter, the battery ESS container, the PCS, and the at least one current translation unit, respectively.

Abstract: A power supply system with UPS, PCS and circuit diagnosis capabilities is disclosed, including: a DC-bus connected to a voltage/current (V/I) meter, a battery energy storage system (ESS) container, a power conditioning system (PCS), at least one current translation unit, and an energy management controller (EMC), wherein, the V/I meter is used to monitor the voltage and current of the DC-bus, and the PCS performs bi-directional conversion between the DC current from the DC-bus and the AC power from an external distribution panel, the current translation unit translate the DC current from the DC-bus into a voltage suitable for at least a critical load, and the EMC controls the operation of the V/I meter, the battery ESS container, the PCS, and the at least one current translation unit, respectively.

Abstract: A ship navigation system includes a memory and a processor. The memory stores instructions. The processor is configured to access the instructions to perform the following: receiving a start point and a terminal point of a target ship in a marine area; receiving real-time environmental information of the marine area; receiving historical shipping lanes of the marine area and historical environmental information corresponding to the historical shipping lanes and; comparing historical environmental information and real-time environmental information to select a target lane leading from a location near the start point to a location near the terminal point; and connecting the start point and the terminal point with the target lane to generate an recommended lane for the target ship in the marine area.

Abstract: An alarm system is used for a vehicle including a fuel pump and an engine. The alarm system includes a measuring device, a computing device and a warning device. The measuring device is used to read a current rotation speed of the engine and respectively measures a plurality of running times of a plurality of rotation speed values of the current rotation speed. The computing device calculates a plurality of consuming times of the fuel pump corresponding to the rotation speed values according to the running times of the rotation speed values, respectively. When an accumulating consuming time accumulated by the consuming times reaches a default critical value, the warning device generates a warning message.

Abstract: An alarm system is used for a vehicle including a fuel pump and an engine. The alarm system includes a measuring device, a computing device and a warning device. The measuring device is used to read a current rotation speed of the engine and respectively measures a plurality of running times of a plurality of rotation speed values of the current rotation speed. The computing device calculates a plurality of consuming times of the fuel pump corresponding to the rotation speed values according to the running times of the rotation speed values, respectively. When an accumulating consuming time accumulated by the consuming times reaches a default critical value, the warning device generates a warning message.

Abstract: The disclosure is related to a system and a method for collision avoidance for vehicle. In the method, the system predicts multiple routes of an abnormal vehicle in a period of time according to historical data when a nearby vehicle receives an alert from the abnormal vehicle. A route potential pattern can be created when the system gets the historical data. The system also computes one or more available routes for the nearby vehicle based on its vehicle information. Every available route has its collision risk value. The system finally provides a recommended route with lower collision risk value when it considers a time of the abnormal vehicle reaches its great change, a time of predicting the nearby vehicle meets the range of route potential pattern, and a safe distance there-between.

Abstract: The disclosure is related to a system and a method for collision avoidance for vehicle. In the method, the system predicts multiple routes of an abnormal vehicle in a period of time according to historical data when a nearby vehicle receives an alert from the abnormal vehicle. A route potential pattern can be created when the system gets the historical data. The system also computes one or more available routes for the nearby vehicle based on its vehicle information. Every available route has its collision risk value. The system finally provides a recommended route with lower collision risk value when it considers a time of the abnormal vehicle reaches its great change, a time of predicting the nearby vehicle meets the range of route potential pattern, and a safe distance there-between.

Abstract: The invention provides a dynamically assigning a parking lot method. First, one or a plurality of parking lots located in a set distance is selected according to a destination provided by the car. Next, a time of parking space to be available is estimated according a real-time parking data and a historical parking data of the one or the plurality of parking lots. Then, a waiting time is estimated according the time of parking space to be available and a car number in a line information of the one or the plurality of parking lots. A driving time of the car to the one or the plurality of parking lots according to a position of the car is estimated. Finally, a total waiting time is estimated according the driving time and the waiting time of the one or the plurality of parking lots.

Abstract: A pedestrian detection system of detecting whether there is a pedestrian in a scene, the pedestrian detection system includes an image-capturing module, a preprocessing module, a human detection module, an image-stitching module and a decision module. The image-capturing module is configured for generating a plurality of first detection image data according to a contrast decision result. The preprocessing module is configured for generating a plurality of first image skeleton data according to the first detection image data. The human detection module is configured for generating a plurality of second image skeleton data. The image-stitching module is configured for stitching the plurality of first detection image data to generate at least one third detection image data. The decision module is configured for generating and outputting a detection result according to the third detection image data. A pedestrian detection method is disclosed herein as well.

Abstract: A building texture extracting apparatus and a method thereof are provided, wherein the building texture extracting apparatus comprises a storage unit and a processor. The storage unit is configured to store an aerial image and a panoramic image of a building. There is a coordinate correlation between the aerial image and the panoramic image. The processor defines an edge segment of a building in the aerial image, calculates an edge function according to the edge segment, projects the edge function onto the panoramic image according to the coordinate correlation to derive an edge curve function, decides an edge curve segment according to the edge curve function, captures an image area under the edge curve segment to be a building texture of the building, and stores the building texture in the storage unit.

Abstract: A pedestrian detection system of detecting whether there is a pedestrian in a scene, the pedestrian detection system includes an image-capturing module, a preprocessing module, a human detection module, an image-stitching module and a decision module. The image-capturing module is configured for generating a plurality of first detection image data according to a contrast decision result. The preprocessing module is configured for generating a plurality of first image skeleton data according to the first detection image data. The human detection module is configured for generating a plurality of second image skeleton data. The image-stitching module is configured for stitching the plurality of first detection image data to generate at least one third detection image data. The decision module is configured for generating and outputting a detection result according to the third detection image data. A pedestrian detection method is disclosed herein as well.

Abstract: A building texture extracting apparatus and a method thereof are provided, wherein the building texture extracting apparatus comprises a storage unit and a processor. The storage unit is configured to store an aerial image and a panoramic image of a building. There is a coordinate correlation between the aerial image and the panoramic image.