Abstract: This document describes an accelerator circuitry for facilitating acceleration of non-maximum suppression (NMS) for detection of objections within an image.

Abstract: Disclosed herein is a system and method for reducing false positives in object detection frameworks. A human reviews objects detected by the object detection framework and indicates whether the object is an object of a class for which the object detection framework is trained to detect. When an indication of a false positive been received, a feature representation of displayed object is stored in a gallery. During an inference or testing phase, the gallery is searched for a feature representation matching the feature representation of the detected objects, and, if a match is found, the detected object is deemed to be a false positive and is not identified as an object for which the object detection framework has been trained to detect.

Abstract: Systems and methods for machine-learning based cycle time tracking and reporting for vehicles are provided. A system includes a processor coupled with memory. The system identifies one or more models trained with machine learning relating to physical characteristics of vehicles and location designations associated with vehicle areas. The system receives, from one or more cameras, a video stream that captures a vehicle disposed in a vehicle area comprising a location designation. The system determines, based on an analysis of a plurality of frames of the video stream and via the one or more models, a type of the vehicle disposed in the vehicle area and a duration the vehicle is disposed in the vehicle area. The system performs, based on the type of the vehicle and a comparison of the duration of the vehicle with a threshold, an action to cause delivery of the vehicle from the vehicle area.

Abstract: Systems and methods for reconstructing field lines within video may include one or more processor configured by machine-readable instructions. Exemplary implementations may include identifying a plurality of candidate field lines in one or more frame of video; training a neural network to detect one or more field object; detecting, using the neural network, one or more field object in the one or more fame of video; assigning, using the detected one or more field object, a confidence value to each of the plurality of candidate field lines; removing one or more low confidence candidate field line from the plurality of candidate field lines; and retaining one or more high confidence candidate field line from the plurality of candidate field lines.

Abstract: A weed spot-spraying system is described for carrying out a spot-based weed spraying method based upon a classification value rendered from a sub-field image in accordance with a machine learning-based trained model applied by a processor to the sub-field image. The system includes a camera; a spray nozzle assembly including a spray nozzle; and a processor. The method carried out by the system includes acquiring, by the camera, a full field of view image of a crop floor. The method further includes extracting, from the full field of view image, a sub-field image corresponding to the spray nozzle positioned to provide a spray field extending over a part of the crop floor depicted in the sub-field image; rendering, by the processor in accordance with the machine learning-based trained model, a classification for the sub-field image; and selectively activating the spray nozzle in accordance with the classification for the sub-field image.

Abstract: Disclosed are a method for accelerating supervised training of a spiking neural network. The method includes measuring first and second membrane potentials for each time step during a training process, extracting distribution data of the first and second membrane potentials based on the first and second membrane potentials for the each time step, calculating a threshold value to be used in a subsequent training process based on the distribution data of the first and second membrane potentials, classifying images having no training contribution based on the threshold value calculated in a previous training process, and terminating the training at the time step based on determining that the image does not have the training contribution when a difference between the first and second membrane potentials in the time step is greater than the threshold value.

Abstract: A computer-implemented method for ascertaining an optimal architecture for a neural network that solves a given task in accordance with given boundary conditions and/or optimization goals. The method includes: providing a graph of the possible architectures of nodes and edges, wherein nodes correspond to data, edges correspond to parameterized operations to be carried out on the data, and a path which traverses the entire graph corresponds to an architecture; in a search phase, generating candidate architectures based on already known architectures, wherein the candidate architectures are similar but not identical to the known architectures in accordance with a predetermined criterion; evaluating the candidate architectures using the given boundary conditions and/or optimization goals; ascertaining a candidate architecture having the best rating as the sought optimal architecture.

Abstract: A system and methods are provided for generating photogrammetry data acquisition plans of target structures, including: establishing a location and orientation of the target structure; identifying on the target structure predetermined target objects designated for data acquisition; correlating shapes, locations, and orientations of the target objects with the shape, location, and orientation of the target structure; and applying the shapes, locations, and orientations of the target structure and of the target objects, to generate a data acquisition plan for drone photogrammetry of the target structure and of the target objects.

Abstract: A surgical instrument navigation system and method of use is provided that visually simulates a virtual volumetric scene of a body cavity of a patient from a point of view of a surgical instrument residing in the cavity of the patient, wherein the surgical instrument, as provided, may be a steerable surgical catheter with a biopsy device and/or a surgical catheter with a side-exiting medical instrument, among others. Additionally, systems, methods and devices are provided for forming a respiratory-gated point cloud of a patient's respiratory system and for placing a localization element in an organ of a patient.

Abstract: Provided are a damage determination information system, a server device, a terminal apparatus, and a program that can realize at least one of the improvement of the convincing feeling for a damage determination result, the reduction of an investigation cost, or the decrease of fluctuation of the damage determination results by a plurality of investigators. A provisional damage determination result is acquired, and a server device extracts a plurality of disaster images related to the provisional damage determination result from a disaster image database (15) and displays the extracted disaster image on a display device of a terminal apparatus. Input of an instruction for settling the damage determination result is received from the terminal apparatus that displays the disaster image, and the settled damage determination result is registered in a damage investigation result database (17).

Abstract: A system trains a model to infer an intent of an entity. The model includes one or more sensors to obtain frames of data, one or more processors, and a memory storing instructions that, when executed by the one or more processors, cause the system to perform steps. A first step includes determining, in each frame of the frames, one or more bounding regions, each of the bounding regions enclosing an entity. A second step includes identifying a common entity, the common entity being present in bounding regions corresponding to a plurality of the frames. A third step includes associating the common entity across the frames. A fourth step includes training a model to infer an intent of the common entity based on data outside of the bounding regions.

Abstract: A model learning device determines a first machine learning model so as to further increase a combined loss function obtained by combining: a first loss function indicating the level of change in the reliability of a second image feature in a feature region of a reconstructed image, from the reliability of a first image feature in a feature region of the original image; and a second loss function indicating the level of recognition error. In addition, the model learning device: determines, so as to further reduce the combined loss function, respective parameter sets for a second machine learning model used in the generation of compressed data, and a third machine learning model used in the generation of the reconstructed image from the compressed data; and determines a parameter set for the fourth machine learning model in common with that for the first machine learning model.

Abstract: Various examples described herein include various mechanisms, techniques, and methods to subtract collected signals caused by a periodic pattern formed on substrates to enable a higher level of defect detection on substrates. Signals detected by various types of metrology and substrate-inspection systems that are caused by periodic patterns on inspected substrates can be reduced or eliminated by, for example, a Fourier analysis of the detected signals. Both gray-scale value thresholds and area thresholds may be applied after the Fourier analysis of the image and are sufficient for defect detection on an image with a substantially-reduced number of false defects or no defects being present in a final image produced after processing. Other techniques and methods are also disclosed.

Abstract: A method of training an image captioning model includes: extracting a feature of a first image from the first image and extracting a feature of a second image from the second image; by encoding viewpoint information based on the feature of the first image and the feature of the second image, obtaining a first image's feature including viewpoint information and a second image's feature including the viewpoint information; obtaining a first image's viewpoint-aligned feature and a second image's viewpoint-aligned feature, based on the first image's feature including the viewpoint information and the second image's feature including the viewpoint information; and generating a caption describing a difference between the first image and the second image, based on the first image's viewpoint-aligned feature and the second image's viewpoint-aligned feature.

Abstract: In an embodiment an apparatus includes a processor configured to generate a feature extraction module using a dataset in which an attribute for each object is defined, receive an image obtained by a camera, extract an attribute of an object of interest from the image using the learned feature extraction module, identify an object re-identification candidate group based on the extracted attribute of the object of interest and re-identify the object of interest based on the identified object re-identification candidate group.

Abstract: A method for the chronological correction of multimodal data includes: receiving a first data set from a reference sensor with measurements at different measurement timepoints, receiving a second data set of a second sensor with measurements at different measurement timepoints, each not exactly matching those of the reference sensor, reading the first and the second data sets by a neural network and identifying a respective plurality of feature vectors for the first and second data set at the respective measurement timepoints, merging and comparing the respective feature vectors, which refer to corresponding, not exactly matching measurement timepoints, by the neural network so that parameters of a chronological correction are identified, and identifying a chronological offset between the respective measurement timepoints of the reference sensor and the second sensor, and/or a corrected data set from the second sensor based on the measurement timepoints of the reference sensor.

Abstract: Implementations are described herein for improving unsupervised domain adaptation (UDA) by using improved adaptive teacher for object detection with cross-domain mix-up. In various implementations, cross-domain training of an object detection machine learning model may include: performing weak augmentation on images from a target domain DT to generate a first set of weakly augmented target domain images; perform strong augmentation on images from the source domain DS and images from the target domain DT to generate a second set of strongly augmented images; processing the second set of strongly augmented images to generate a third set of inter-domain mixes of the images from DS and DT; and jointly train the object detection machine learning model, as a student machine learning model, with a teacher machine learning model using the first and third sets.

Abstract: In an approach for improving operational efficiencies in IoT agricultural ecosystems, a processor receives a request from a user for a production plan to modify one or more values of a set of values of an area of interest. A processor maps the area of interest using a set of data originating from a LIDAR instrument to create a digital elevation model of the area of interest. A processor creates a knowledge set to train a decision engine. A processor calculates one or more modifications to be made to the one or more values of the set of values of the area of interest. A processor generates a map of applicable land options that the user may interact with to select the one or more modifications to be made to the one or more values of the set of values. A processor plots a finalized rendering of the production plan.

Abstract: A segmentation model is trained with an image reconstruction model that shares an encoding. During application of the segmentation model, the segmentation model may use the encoding and network layers trained for the segmentation without the image reconstruction model. The image reconstruction model may include a probabilistic representation of the image that represents the image based on a probability distribution. When training the model, the encoding layers of the model use a loss function including an error term from the segmentation model and from the autoencoder model. The image reconstruction model thus regularizes the encoding layers and improves modeling results and prevents overfitting, particularly for small training sizes.

Abstract: Some aspects of the present invention may include systems and methods of a detecting whether a first image contains a region that has been manipulated, methods comprising obtaining a second image, wherein the second image comprises at least a part of the first image, said at least a part of the first image containing the region suspected of being manipulated; determining a numerical value of an order parameter (S or S2) of the second image; determining a numerical value of an order parameter (S or S2) of a third image, the third image comprising the second image with the region suspected of being manipulated removed; and comparing the numerical value of the second image (S or S2) with the numerical value (S or S2) of the third image to determine if the first image has been altered, by reference to a predefined criteria indicative of a manipulated image.