A method (100), a device (600;700) and a system (800) for processing image data representing a scene for extracting features related to objects in the scene using a convolutional neural network are disclosed. Two or more portions of the image data representing a respective one of two or more portions of the scene are processed (S110), by means of a respective one of two or more circuitries, through a first number of layers of the convolutional neural network to form two or more outputs, wherein the two or more portions of the scene are partially overlapping. The two or more outputs are combined (S120) to form a combined output, and the combined output is processed (S130) through a second number of layers of the convolutional neural network by means of one of the two or more circuitries for extracting features related to objects in the scene.

A method for determining a semantic free space in an environment of a vehicle comprises capturing a two dimensional visual image from the environment of the vehicle via a camera and determining a limitation of a free space within the visual image. Via a sensor, distance data of objects are captured and assigned to the visual image, and the limitation of the free space is transferred to a bird's-eye view based on the assigned distance data. For objects identified in the visual image a respective bounding box and a respective classification are determined. Objects limiting the free space are selected, and their bounding box is assigned to the limitation of the free space in the bird's-eye view. Finally, segments of the limitation of the free space are classified according to the classification of each bounding box of the selected objects.

Described herein are systems, methods, and non-transitory computer readable media for using 3D point cloud data such as that captured by a LiDAR as ground truth data for training an instance segmentation deep learning model. 3D point cloud data captured by a LiDAR can be projected on a 2D image captured by a camera and provided as input to a 2D instance segmentation model. 2D sparse instance segmentation masks may be generated from the 2D image with the projected 3D data points. These 2D sparse masks can be used to propagate loss during training of the model. Generation and use of the 2D image data with the projected 3D data points as well as the 2D sparse instance segmentation masks for training the instance segmentation model obviates the need to generate and use actual instance segmentation data for training, thereby providing an improved technique for training an instance segmentation model.

A target detection apparatus is configured to recognize a stationary object present within a detection range of an external sensor based on map information. Next, the target detection apparatus is configured to determine, by checking an image of the stationary object detected by the external sensor against the stationary object recognized from the map information, whether the image of the stationary object detected by the external sensor includes an undetected region. When the undetected region is identified, the target detection apparatus is configured to recognize an undetectable object present between the stationary object and a vehicle.

An optimal automatic mapping method between a real image and a thermal image in a body heat tester, and the body heat tester using the method. The real image from the real imaging camera has wider angle of view than the thermal image from the thermal imaging camera, to maximize the use of thermal imaging without omission of thermal imaging pixels in a thermal inspection device using an infrared imaging device. The body heat tester comprises a thermal imaging camera, a real imaging camera and a data processing unit. The data processing unit matches the thermal and real images, obtains the reconstructed real image matched with the thermal image by stretching or shortening the top, bottom, left, and right of the real image based on the thermal image, and detects the body heat (temperature) of the subject using the thermal image and the reconstructed real image.

A system for generation of training dataset is provided. The system controls a depth sensor to capture, from a first viewpoint, a first image a first depth value associated with the first object. The system receives tracking information from a handheld device associated with the depth sensor, based on a movement of the handheld device and the depth sensor in a 3D space. The system generates graphic information corresponding to the first object based on the received tracking information. The graphic information includes the first object from a second viewpoint. The system calculates a second depth value associated with the first object, based on the graphic information. The system generates, for a neural network model, a training dataset which includes a first combination of the first image and the first depth value, and a second combination of second images corresponding to the graphic information and the second depth value.

A system and method for analyzing risk and providing risk mitigation instructions. The system receives analyzes sensor data and other data corresponding to a user to determine a test group. The system uses the test group to determine a risk score, and, subsequently, a risk mitigation strategy. Machine learning techniques are implemented to refine how the test group, risk score, and mitigation are each selected.

An image processing method and apparatus, a device, a video processing method and a storage medium are provided. The image processing method includes: receiving an input image; and processing the input image by using the convolutional neural network to obtain an output image. A definition of the output image is higher than a definition of the input image. Processing the input image by using the convolutional neural network to obtain the output image includes: performing feature extraction on the input image; concatenating the input image and the plurality of first images; performing the feature extraction on the first image group; fusing the plurality of second images and the plurality of first images; concatenating the input image and the plurality of third images to obtain a second image group; and performing the feature extraction on the second image group to obtain the output image.

A method includes receiving an input specifying a recognition target. The method further includes selecting a plurality of models of an initial recognition layer based on the recognition target, and selecting a plurality of models of a final recognition layer based on the recognition target. The method includes obtaining sensor data from two or more sensors of a plurality of sensors, providing the sensor data to the plurality of models of the initial recognition layer to obtain an initial set of identifications, providing sensor data to the plurality of models of the final recognition layer to obtain a final set of identifications, and outputting an identification from at least one of the initial set of identifications or the final set of identifications.

Implementations are described herein for estimating aggregate trait values for agricultural plots. In various implementations, a plurality of images depicting crops in an agricultural plot may be processed using a feature extraction portion of aggregate trait estimation model to generate a corresponding plurality of embeddings. The plurality of embeddings may be combined to obtain a unified representation of the crops in the agricultural plot. Using a prediction portion of the aggregate trait estimation model, the unified representation of the crops may be processed to estimate, as a direct output of the aggregate trait estimation model, an aggregate trait value of the crops in the agricultural plot. Data indicative of the estimated aggregate trait value of the crops in the agricultural plot may be output at a computing device.