In accordance with various embodiments of the disclosed subject matter, a method and a system for vision-centric deep-learning-based road situation analysis are provided. The method can include: receiving real-time traffic environment visual input from a camera; determining, using a ROLO engine, at least one initial region of interest from the real-time traffic environment visual input by using a CNN training method; verifying the at least one initial region of interest to determine if a detected object in the at least one initial region of interest is a candidate object to be tracked; using LSTMs to track the detected object based on the real-time traffic environment visual input, and predicting a future status of the detected object by using the CNN training method; and determining if a warning signal is to be presented to a driver of a vehicle based on the predicted future status of the detected object.

A method for processing sound used in a speech recognition robot is disclosed. The method for processing sound comprises the steps of: recognizing, by a robot, an obstacle on a driving path; calculating, by the robot, a driving distance to the obstacle; calculating a driving speed, by the robot; and determining, by the robot, a point in time at which a transient sound is generated by an impact caused by passing through the obstacle, wherein the point in time at which the transient sound is generated may be determined, by the robot, from the driving distance to the obstacle and the driving speed. The robot can transmit and receive a wireless signal on a mobile communication network established according to 5G (fifth generation) communication.

An object detection device is configured to be mounted to an own vehicle to detect an object present around the own vehicle. The object detection device includes a coordinate acquisition section and an object recognition section. The coordinate acquisition section acquires position coordinates of a detection point corresponding to the object on the basis of a plurality of images captured at different positions along with movement of the own vehicle by one imaging section mounted to the own vehicle. The object recognition section recognizes the object without using position coordinates of a movement point that is a detection point corresponding to a moving object but using position coordinates of a stop point that is a detection point different from the movement point.

An optical system includes, in order from an enlargement conjugate side to a reduction conjugate side, a front unit including a plurality of lenses, an aperture stop, and a rear unit including a plurality of lenses. A projection characteristic of the optical system representing a relationship between a half angle of view and an image height on an image plane satisfies a predetermined condition.

A computer-implemented method is provided for action localization. The method includes converting one or more video frames into person keypoints and object keypoints. The method further includes embedding position, timestamp, instance, and type information with the person keypoints and object keypoints to obtain keypoint embeddings. The method also includes predicting, by a hierarchical transformer encoder using the keypoint embeddings, human actions and bounding box information of when and where the human actions occur in the one or more video frames.

An optical system includes a first lens disposed closest to an enlargement conjugate position, a second lens adjacent to the first lens, a diaphragm disposed closer to a reduction conjugate position than the second lens, and a final lens disposed closest to a reduction conjugate position. Each of the first lens, the second lens, and the final lens has an aspherical surface. The aspherical surface of each of the first and second lenses has an inflection point. An imaging magnification is different between a first area of the optical system and a second area on a periphery side of the first area.

A method for identifying and/or tracking objects in a spatial area. The method includes observing the area with the aid of at least one event-based sensor, the event-based sensor including light-sensitive pixels, and a relative change of the light intensity incident upon a pixel by at least a predefined percentage prompting the sensor to output an event assigned to this pixel. The method further includes, in response to the sensor, outputting a new event, ascertaining an assessment for this event which is a measure of the extent to which this event matches an already detected distribution of events, and/or of the extent to which it is plausible that the event stems from an already detected object; and in response to the ascertained assessment meeting a predefined criterion, assigning the new event to the already detected distribution, or the already detected object.

A method for detecting objects and labeling the objects with distances in an image includes steps of: obtaining a thermal image from a thermal camera, an RGB image from an RGB camera, and radar information from an mmWave radar; adjusting the thermal image based on the RGB image to generate an adjusted thermal image, and generating a fused image based on the RGB image and the adjusted thermal image; generating a second fused image based on the fused image and the radar information; detecting objects in the images, and generating, based on the fused image, another fused image including bounding boxes marking the objects; and determining motion parameters of the objects.

Apparatuses, systems, and techniques are presented to remove objects from images and perform inpainting for regions of object removal. In at least one embodiment, one or more neural networks are used to remove one or more objects from one or more images, wherein the one or more objects are of a similar type.

The present disclosure generally relates to a system of a delivery device for combining sensor data from various types of sensors to generate a map that enables the delivery device to navigate from a first location to a second location to deliver an item to the second location. The system obtains data from RGB, LIDAR, and depth sensors and combines this sensor data according to various algorithms to detect objects in an environment of the delivery device, generate point cloud and pose information associated with the detected objects, and generates object boundary data for the detected objects. The system further identifies object states for the detected object and generates the map for the environment based on the detected object, the generated object proposal data, the labeled point cloud data, and the object states. The generated map may be provided to other systems to navigate the delivery device.