Abstract: A display control device includes: an image acquisition unit configured to acquire captured image data from an imaging unit that captures an image of a peripheral area of a vehicle; a display control unit configured to cause image data based on the captured image data to be displayed on a screen; and an operation receiving unit configured to receive designation of an arbitrary point on the screen, in which the display control unit enlarges display on the screen about the designated point as a center of enlargement in a state where a display position of the designated point does not move on the screen when the operation receiving unit receives the designation of the arbitrary point on the screen.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection policy neural network, wherein the action selection policy neural network is configured to process an observation characterizing a state of an environment to generate an action selection policy output, wherein the action selection policy output is used to select an action to be performed by an agent interacting with an environment. In one aspect, a method comprises: obtaining an observation characterizing a state of the environment subsequent to the agent performing a selected action; generating a latent representation of the observation; processing the latent representation of the observation using a discriminator neural network to generate an imitation score; determining a reward from the imitation score; and adjusting the current values of the action selection policy neural network parameters based on the reward using a reinforcement learning training technique.

Abstract: A method includes obtaining a reference image and a target image each representing an environment containing moving features and static features. The method also includes determining an object mask configured to mask out the moving features and preserves the static features in the target image. The method additionally includes determining, based on motion parallax between the reference image and the target image, a static depth image representing depth values of the static features in the target image. The method further includes generating, by way of a machine learning model, a dynamic depth image representing depth values of both the static features and the moving features in the target image. The model is trained to generate the dynamic depth image by determining depth values of at least the moving features based on the target image, the object mask, and the static depth image.

Abstract: A system may generate image data of users within a facility. Such users may be part of a group that is unknown to the system. The system can predict group data for two or more users based on a resemblance or the users being within a threshold distance of each other. However, if a confidence level associated with the predicted group data is below a threshold value, the group data is deemed unreliable and assistance from an associate is deemed necessary. A user interface that includes a portion of the image data, information about the predicted group data, and other interface elements is presented to the associate via a display. Based on the input data received from the associate, the group data can be confirmed or rejected. If the group data is confirmed, an association is made between the users and a group identifier.

Abstract: A driver monitoring system includes at least two cameras configured to record images of a driver within a cabin of a vehicle. A first camera is disposed on a driver side of the cabin and a second camera is disposed on a passenger side of the cabin. At least one of the at least two cameras is disposed in a mirror replacement display or in a mirror replacement display location. A controller is in communication with the at least two cameras and is configured to determine a pose of the driver based on images from the at least two cameras. At least one of the images originates from at least one of the first camera and the second camera.

Abstract: A method of determining ripeness of a growing mushroom involves obtaining a sequence of images of a growing mushroom over a period of time and, from the sequence of images, measuring one or more features of the growing mushroom to obtain temporal measurements for the one or more features. The temporal measurements indicate rates at which the one or more features are changing. The temporal measurements are compared to pre-determined ripeness functions to determine the ripeness of the mushroom. The mushroom is picked if the comparison indicates that the mushroom is ready to be harvested.

Abstract: In various examples, surface profile estimation and bump detection may be performed based on a three-dimensional (3D) point cloud. The 3D point cloud may be filtered in view of a portion of an environment including drivable free-space, and within a threshold height to factor out other objects or obstacles other than a driving surface and protuberances thereon. The 3D point cloud may be analyzed—e.g., using a sliding window of bounding shapes along a longitudinal or other heading direction—to determine one-dimensional (1D) signal profiles corresponding to heights along the driving surface. The profile itself may be used by a vehicle—e.g., an autonomous or semi-autonomous vehicle—to help in navigating the environment, and/or the profile may be used to detect bumps, humps, and/or other protuberances along the driving surface, in addition to a location, orientation, and geometry thereof.

Abstract: A mobile device can implement a neural network-based domain transfer scheme to modify an image in a first domain appearance to a second domain appearance. The domain transfer scheme can be configured to detect an object in the image, apply an effect to the image, and blend the image using color space adjustments and blending schemes to generate a realistic result image. The domain transfer scheme can further be configured to efficiently execute on the constrained device by removing operational layers based on resources available on the mobile device.

Abstract: A method, apparatus, system, and computer program product for monitoring a platform. Images are received by a computer system from a sensor system positioned to monitor movement of human beings relative to the platform. A set of the human beings is identified by the computer system in the images. The movement of the set of the human beings relative to the platform is determined by the computer system using the images. A count of the human beings on the platform is determined by the computer system based on the movement determined for the set of the human beings. A set of actions is performed by the computer system based on the count of the human beings on the platform.

Abstract: Techniques and mechanisms for processing differential video data with a spiking neural network to provide action recognition functionality. In an embodiment, the spiking neural network is coupled to receive and process a first one or more spike trains which represent an encoded version of a sequence comprising frames of differential video data. In turn, the frames of differential video data are each based on a difference between a respective two frames of raw video data. Based on the processing of the first one or more spike trains, the spiking neural network may output a second one or more spike trains. In another embodiment, the second one or more spike trains are provided to train the spiked neural network to recognize an activity type, or to classify a video sequence as including a representation of an instance of the activity type.

Abstract: Triangulation is applied, by a method and a device, to determine a scene point location in a global coordinate system, GCS, based on location coordinates for observed image points in plural views of a scene, and global-local transformation matrices for the views. The local coordinates are given in local coordinate systems, LCSs, arranged to define distance coordinate axes that are perpendicular to image planes of the views. The scene point location is determined by minimizing a plurality of differences between first and second estimates of the scene point location in the LCSs, the first estimates being given by linear scaling of the location coordinates, and the second estimate being given by operating the transformation matrices on the scene point location. An improved robustness to observation errors is achieved by defining the plurality of differences to penalize differences that includes the linear scaling as applied along the distance coordinate axes.

Abstract: A method of compressing image data comprising a set of image values each representing a position in image-value space so as to define an occupied region thereof. The method comprises selectively applying a series of compression transforms to subsets of the image data items to generate a transformed set of image data items occupying a compacted region of value space. The method further comprises identifying a set of one or more reference data items that quantizes the compacted region in value space. For each image data item in the set of image data items, a sequence of decompression transforms from a fixed set of decompression transforms is identified that generates an approximation of that image data item when applied to a selected one of the one or more reference data items. Each image data item in the set of image data items is encoded as a representation of the identified sequence of decompression transforms for that image data item.

Abstract: A system, and method of operating the same detects moving targets in images and performs image turbulence correction. The system includes an automatic target recognizer (ATR) system including a database. The ATR includes a feature extractor and processor arranged to detect a plurality of reference features associated with targets within image frames, and calculate a position of the plurality of reference features. The system includes an image processor arranged to receive the position, demosaic the image frames into a plurality of video tiles, iteratively process the video tiles for turbulence correction to generate turbulence corrected video tiles associated with acquired targets; convert the turbulence corrected video tiles into a single video frame tile including turbulence degradation correction; and mosaic each of the single video frame tiles to generate a full field of view turbulence corrected video stream.

Abstract: The present disclosure may provide a method for perfusion analysis. The method may include: obtaining a plurality of scan images corresponding to a plurality of time points; obtaining a plurality of time-density discrete points based on the plurality of scan images; determining an initial time-density curve based on the plurality of time-density discrete points, the initial time-density curve indicating a density variation of a contrast agent in an organ or tissue over time, the organ or tissue corresponding to a pixel or voxel in the plurality of scan images; obtaining a first perfusion model; determining a first perfusion parameter based on the first perfusion model and the initial time-density curve; obtaining a second perfusion model; and determining a second perfusion parameter based on the second perfusion model and the first perfusion parameter.

Abstract: Devices and techniques are generally described for articulated three-dimensional pose tracking. In some examples, a plurality of frames of image data captured by one or more cameras may be received. First feature data representing the plurality of frames of image data may be determined using a backbone network. The first feature data may be projected into three-dimensional (3D) space. In some examples, 3D location data describing respective 3D locations of one or more persons represented by the first feature data projected in the 3D space may be determined. The first feature data and the 3D location data may be sent to a four-dimensional (4D) convolutional neural network (CNN). The 4D CNN may generate second feature data comprising respective 3D representations of the one or more persons. Three dimensional pose data representing articulated 3D pose information for the one or more persons may be generated.

Abstract: There are provided an information suggestion system, an information suggestion method, a program, and a recording medium capable of suggesting, to a user, information on a product matching a hobby and the preference of the user from an image group of the user. In the information suggestion system, the information suggestion method, the program, and the recording medium, an image group acquisition unit acquires an image group of a user, and an image analysis unit detects an attribute of each image. An imaging number count unit counts an imaging number of same-attribute images, and an imaging frequency calculation unit calculates an imaging frequency of the same-attribute images. An imaging purpose estimation unit estimates an imaging purpose of the same-attribute images.

Abstract: The described embodiments relate to systems, methods, and apparatus for providing a multimodal deep memory network (200) capable of generating patient diagnoses (222). The multimodal deep memory network can employ different neural networks, such as a recurrent neural network and a convolution neural network, for creating embeddings (204, 214, 216) from medical images (212) and electronic health records (206). Connections between the input embeddings (204) and diagnoses embeddings (222) can be based on an amount of attention that was given to the images and electronic health records when creating a particular diagnosis. For instance, the amount of attention can be characterized by data (110) that is generated based on sensors that monitor eye movements of clinicians observing the medical images and electronic health records.

Abstract: Physical and logical components of an apparatuses, systems and methods for and related to detecting, identifying, and categorizing construction site objects and other objects on real property (“objects”) through artificial intelligence machine learning analysis of object sensor data to identify an object in the object sensor data, to determine a categorization of the object, to determine at least one of a site map, a hazardous condition, a theft, and or a behavior of the objects and to output warnings and utilization reports with respect to equipment, vehicles, personnel.

Abstract: The present disclosure is directed toward systems and methods to quickly and accurately identify boundaries of a displayed document in a live camera image feed, and provide a document boundary indicator within the live camera image feed. For example, systems and methods described herein utilize different display document detection processes in parallel to generate and provide a document boundary indicator that accurately corresponds with a displayed document within a live camera image feed. Thus, a user of the mobile computing device can easily see whether the document identification system has correctly identified the displayed document within the camera viewfinder feed.

Abstract: A method, apparatus, and system for identifying objects based on grayscale images in the operation of an autonomous driving vehicle is disclosed. In one embodiment, one or more first color images are received from a camera mounted at an autonomous driving vehicle (ADV). One or more first grayscale images are generated based on the one or more first color images, which comprises converting each of the one or more first color images into one of the first grayscale images. One or more objects in the one or more first grayscale images are identified based on a pre-trained grayscale perception model. A trajectory for the ADV is planned based at least in part on the identified one or more objects. Control signals are generated to drive the ADV based on the planned trajectory.