Abstract: A method for detecting a grasping position of a robot in grasping a target object includes: collecting a target RGB image and a target Depth image of the target object at different view angles; inputting each of the target RGB image to a target object segmentation network for calculation to obtain an RGB pixel region of the target object in the target RGB image and a Depth pixel region of the target object; inputting the RGB pixel region to an optimal grasping position generation network to obtain an optimal grasping position for grasping the target object; inputting the Depth pixel region of the target object and the optimal grasping position to a grasping position quality evaluation network to calculate a score of the optimal grasping position; and selecting an optimal grasping position corresponding to a highest score as a global optimal grasping position of the robot.

Abstract: A cloud-based trajectory map generation method, device, apparatus and application. The method includes: acquiring first trajectory map data and second trajectory map data; performing a map-initializing treatment to align the two trajectory map data to obtain an initial value of a transformation relationship therebetween; interpolating the two trajectory map data to obtain corresponding data at preset interpolation points thereto; determining a residual equation based on the corresponding data at the preset interpolation points and the initial value of the transformation relationship to obtain an optimal solution of the transformation relationship based thereon; and generating a trajectory map based on the optimal solution of the transformation relationship.

Abstract: A panum's area measurement method includes: projecting a first parallax image of a spatial object to the left eye of a user under test, and projecting a second parallax image of the spatial object to the right eye of the user under test, the first parallax image comprising a first homologous point and the second parallax image comprising a second homologous point; adjusting a horizontal parallax amount between the first homologous point and the second homologous point until the user under test observes the spatial object producing a ghost; acquiring a parallax amount parameter ?ne; calculating a horizontal physical spacing ?x between the first homologous point and the second homologous point based on the parallax amount parameter ?ne; and calculating a panum's area range (?in, ?out) of the user under test based on the horizontal physical spacing ?x.

Abstract: A binocular brightness sensitivity measurement method based on a wearable display device includes: loading a first test image and a second test image which has a brightness different from that of the first test image respectively for two eyes of a user under test; adjusting the brightness of the first test image and/or the second test image until a predefined brightness sensitivity perception test condition is fulfilled; acquiring a brightness difference between the first test image and the second test image; and determining a binocular brightness sensitivity of the user under test based on the brightness difference.

Abstract: An information processing method, device, cloud-based processing device, and computer program product are related to the field of data processing technologies and can cause an increased efficiency in detecting whether a road area contains a depression region. The information processing method includes: acquiring a depth image; processing the depth image to obtain a means-by-row graph, based on which a road area in the depth image is determined; determining a suspected depression region in the road area; and judging over the suspected depression region based on a depression threshold to determine whether the depth image contains a depression region.

Abstract: A method, a device and a terminal for generating training data is provided. The method for generating training data includes: obtaining an original image; determining a transferred image based on the image style transfer model and the original image, wherein the image style transfer model is obtained by minimizing a loss function, the loss function is determined by the original loss function the background loss function and the foreground loss function; determining the training data based on the transferred image. The difference between the generated training data and the target image is small, thereby improving the accuracy of the training model obtained by using the training data.

Abstract: An image annotating method includes: acquiring an image collected at a terminal; acquiring voice information associated with the image; annotating the image according to the voice information; and storing an annotated result of the image.

Abstract: A method for generating training data includes: defining a corresponding category tag for an object model in a three-dimensional scenario; acquiring a plurality of corresponding scenario images by modifying scenario parameters of the three-dimensional scenario; rendering the object model to a monochromic material corresponding to the category tag of the object model; acquiring a semantic segmentation image corresponding to each scenario image based on the rendered object model; and storing each scenario image and the semantic segmentation image corresponding to the scenario image as a set of training data.

Abstract: A method for training an object identification neural network based on a distorted fisheye image includes: acquiring the distorted fisheye image, wherein the distorted fisheye image comprises at least one target object; dividing the distorted fisheye image into a plurality of zones according to a distortion degree; determining a zone to which each of the at least one target object belongs; categorizing the at least one target object according to the zone to which each of the at least one target object belongs; and inputting the distorted fisheye image and a category to which each of the at least one target object belongs into the object identification neural network to train the object identification neural network.

Abstract: The embodiment of the present disclosure provides an image identification system and an image identification method, and relates to the field of identification technology for improving the classification accuracy of the existing image classification model.

Abstract: A method for improving robustness of a visual-inertial navigation system includes: determining a reference frame; correcting the reference frame; and performing non-linear optimization for frames other than the reference frame and an oldest frame according to the oldest frame and the corrected reference frame.

Abstract: The embodiment of the present invention provides a navigation method and apparatus, and a terminal device, relating to the technical field of navigation, and for reducing or avoiding the impact of network delay on real-time obstacle detection and avoidance. The method includes: detecting an obstacle to obtain first obstacle information; obtaining scene information and sending the scene information to a remote server, so that the remote server obtains second obstacle information according to the scene information, wherein the accuracy of the second obstacle information is greater than the accuracy of the first obstacle information; and if the second obstacle information sent by the remote server is not received, avoiding the obstacle according to the first obstacle information. The embodiment of the present invention is applied to navigation.

Abstract: A method and apparatus for controlling a device to move, a storage medium, and an electronic device. The method includes: collecting a first RGB-D image of a surrounding environment of a target device according to a preset period when the target device moves; obtaining a second RGB-D image of a preset number of frames from the first RGB-D image; obtaining a pre-trained deep Q network model DQN training model, and performing migration training on the DQN training model according to the second RGB-D image to obtain a target DQN model; obtaining a target RGB-D image of the current surrounding environment of the target device; inputting the target RGB-D image into the target DQN model to obtain a target output parameter, and determining a target control strategy according to the target output parameter; and controlling the target device to move according to the target control strategy.

Abstract: A target identification method includes: using information of a to-be-detected target acquired within a predetermined time period as judgment information; acquiring an identification result of the to-be-detected target at a current time and outputting the identification result; judging whether the attribute type corresponding to the identification result is an attribute type having the highest priority; and if the attribute type corresponding to the identification result is not the attribute type having the highest priority, using information of the to-be-detected target acquired within a next predetermined time period as the judgment information, and returning to the step of acquiring an identification result of the to-be-detected target at a current time and outputting the identification result.

Abstract: An obstacle avoidance reminding method includes: performing ground detection based on acquired image data to acquire ground information of a road; performing passability detection based on the acquired ground information, and determining a traffic state of the road; if it is determined that the road is impassable, performing road condition detection for the road to acquire a first detection result, and performing obstacle detection for the road to acquire a second detection result; and determining obstacle avoidance reminding information based on the first detection result and the second detection result.

Abstract: An obstacle detection method and apparatus are provided. The obstacle detection method provided includes: obtaining a to-be-detected image; determining a road surface area and a non-road surface area in the to-be-detected image according to pixel information contained in the to-be-detected image; respectively determining an outermost layer contour line of the road surface area and a contour line of the non-road surface area; and when the contour line of at least one non-road surface area is located in the area contained in the outermost layer contour line of the road surface area, determining a physical object contained in the at least one non-road surface area as an obstacle. The present application is applied to a process of detecting an obstacle.

Abstract: The embodiment of the present disclosure discloses an intelligent blind guide method and apparatus, and relates to the field of artificial intelligence, so as to solve the problem that an intelligent blind guide system requires the whole-process intervention of an artificial customer service staff, which results in high work intensity. The intelligent blind guide method includes: obtaining a confidence of an intelligent blind guide according to sensor information, wherein the confidence indicates the reliability of blind guide information generated according to the sensor information without artificial decision; generating the blind guide information according to the sensor information if the confidence is greater than or equal to a preset threshold, and importing artificial decision to generate the blind guide information if the confidence is less than the preset threshold. The embodiment of the present disclosure is applied to a blind guide helmet.

Abstract: A method for detecting a grasping position of a robot in grasping a target object includes: collecting a target RGB image and a target Depth image of the target object at different view angles; inputting each of the target RGB image to a target object segmentation network for calculation to obtain an RGB pixel region of the target object in the target RGB image and a Depth pixel region of the target object; inputting the RGB pixel region to an optimal grasping position generation network to obtain an optimal grasping position for grasping the target object; inputting the Depth pixel region of the target object and the optimal grasping position to a grasping position quality evaluation network to calculate a score of the optimal grasping position; and selecting an optimal grasping position corresponding to a highest score as a global optimal grasping position of the robot.

Abstract: A method for training an object identification neural network based on a distorted fisheye image includes: acquiring the distorted fisheye image, wherein the distorted fisheye image comprises at least one target object; dividing the distorted fisheye image into a plurality of zones according to a distortion degree; determining a zone to which each of the at least one target object belongs; categorizing the at least one target object according to the zone to which each of the at least one target object belongs; and inputting the distorted fisheye image and a category to which each of the at least one target object belongs into the object identification neural network to train the object identification neural network.

Abstract: The embodiment of the present application provides a method and apparatus for detection of false alarm obstacle, and relates to the field of identification and detection technology, in order to improve accuracy for the obstacle detection. The method includes: obtaining a first view at a moment t and a second view at a moment t?1, collected by a camera; determining location information of the same obstacle to be detected in the first view and the second view; determining motion information of the camera from moment t?1 to the moment t; judging whether the location information of the same obstacle to be detected in two views being matched with the motion information of the camera; and if not, judging the obstacle to be detected as a false alarm obstacle. The embodiment of the present application is applied to obstacle detection.