The method for item identification preferably includes determining visual information for an item; calculating a first encoding using the visual information; calculating a second encoding using the first encoding; determining an item identifier for the item using the second encoding; optionally presenting information associated with the item to a user; and optionally registering a new item.

The method for item identification preferably includes determining visual information for an item; calculating a first encoding using the visual information; calculating a second encoding using the first encoding; determining an item identifier for the item using the second encoding; optionally presenting information associated with the item to a user; and optionally registering a new item.

A method of self-supervised learning for detection network using deep Q-network includes steps of: performing object detection on first unlabeled image through the detection network trained with training database to generate first object detection information and performing learning operation on a first state set corresponding to the first object detection information to generate a Q-value, if an action of the Q-value accepts the first unlabeled image, testing the detection network, retrained with the training database additionally containing a labeled image of the first unlabeled image, to generate a first accuracy, and if the action rejects the first unlabeled image, testing the detection network without retraining, to generate a second accuracy, and storing the first state set, the action, a reward of the first or the second accuracy, and a second state set of a second unlabeled image as transition vector, and training the deep Q-network by using the transition vector.

A processor-implemented motion information estimating method includes: estimating motion information based on at least one set of initial motion information of a vehicle that is obtained from at least one sensor; predicting a plurality of sets of image feature information corresponding to a periphery of the vehicle based on the estimated motion information; obtaining a plurality of sets of detected image feature information detected from an input image obtained from an image sensor and an accuracy of each of the sets of the detected image feature information using a neural network; evaluating a reliability of each of the sets of the detected image feature information by comparing the sets of the predicted image feature information and the sets of the detected image feature information; and correcting the estimated motion information based on at least one set of the sets of the detected image feature information selected based on a result of the evaluating of the reliability and on the accuracy.

A method and an apparatus for asynchronous data fusion, a storage medium and an electronic device are provided. The method includes: obtaining current frame LiDAR data, and determining current frame LiDAR three-dimensional embeddings; determining a previous frame fused hidden state, and performing a temporal fusion process based on the previous frame fused hidden state and the current frame LiDAR three-dimensional embeddings to generate a current frame temporary hidden state and a current frame output result; and obtaining current frame camera data, determining current frame camera three-dimensional embeddings, and generating a current frame fused hidden state based on the current frame camera three-dimensional embeddings and the current frame temporary hidden state. Asynchronous fusion is performed on the current frame LiDAR data and previous frame camera data, which leads to a low processing latency.

The present disclosure relates to an image processing method and apparatus, an electronic device, and a storage medium. The method includes: acquiring at least two target images; determining an attention map of at least one target in each of the at least two target images according to a result of detecting target of each target image, where the attention map indicates the position of one target in a target image; and determining, based on each target image and the attention map of the at least one target in the each target image, a result of association that indicates whether a correspondence exists between at least some of targets in different target images.

A mobility information provision system includes a collector, a mapping unit, an overall generator, and a specific generator. The collector collects information about movement of a plurality of mobile bodies. The mapping unit maps positions of the plurality of mobile bodies on the basis of the information collected by the collector. The overall generator repeatedly performs a first generation process of generating course-related information by using information including the mapped positions. The specific generator performs, in a case where a specific area for one or more mobile bodies, out of the plurality of mobile bodies, has been set, a second generation process of generating the course-related information for the one or more mobile bodies present in the specific area, by using the information including the mapped positions. The specific generator executes the second generation process in precedence over the first generation process executed by the overall generator.

The present application provides a method and apparatus for pose planar constraining on the basis of planar feature extraction, wherein the method includes: inputting the acquired RGB color image and point cloud image into spatial transformation network to obtain two-dimensional and three-dimensional affine transformation matrixes; extracting the planar features of the transformed two-dimensional affine transformation matrix and three-dimensional affine transformation matrix; inputting the acquired planar features into the decoder and obtain the pixel classification of the planar features; clustering the vectors corresponding to the planar pixels to obtain the segmentation result of the planar sample; using planar fitted by the segmentation result to make planar constraint to the pose calculated by vision algorithm. The application combines RGB-D information to perform plane extraction, and designs a new spatial transformation network to transform two-dimensional color image and three-dimensional point cloud image.

A system and method for predicting driving actions based on intent-aware driving models that include receiving at least one image of a driving scene of an ego vehicle. The system and method also include analyzing the at least one image to detect and track dynamic objects located within the driving scene and to detect and identify driving scene characteristics associated with the driving scene and processing an ego-thing graph associated with the dynamic objects and an ego-stuff graph associated with the driving scene characteristics. The system and method further include predicting a driver stimulus action based on a fusion of representations of the ego-thing graph and the ego-stuff graph and a driver intention action based on an intention representation associated with driving intentions of a driver of the ego vehicle.

An apparatus and method for compensating for an error of a vehicle sensor for enhancing performance for identifying the same object are provided. The apparatus includes a rotation angle error calculator that calculates a rotation angle error between sensor object information and sensor fusion object information. A position error calculator calculates a longitudinal and lateral position error between the sensor object information and the sensor fusion object information. A sensor error compensator calculates a sensor error based on the calculated rotation angle and a position error. In calculating the rotation angle error, the sensor error compensator corrects an error of the sensor object information based on the rotation angle error, and compensates for the sensor error based on the longitudinal and lateral position error between the corrected sensor object information and the sensor fusion object information.