Abstract: A method for recognizing an object in a video stream may include receiving a video stream comprising a plurality of video frames from a video source. The method may also select at least one video frame from the video frames according to a frame selection rate. The method may also partition the selected video frame into a first plurality of image blocks, and recognize, out of the first plurality of image blocks, a second plurality of image blocks which comprise an image of an object, the recognition being based on an image recognition parameter determined by a machine-learning algorithm. The method may also determine that at least one of the second plurality of image blocks corresponds to the object based on a likelihood metric, the likelihood metric being determined by the processor based on at least the frame selection rate, and display, on a display, information identifying the object.

Abstract: In some examples, a machine learning model may be trained to denoise an image. In some examples, the machine learning model may identify noise in an image of a sequence based at least in part, on at least one other image of the sequence. In some examples, the machine learning model may include a recurrent neural network. In some examples, the machine learning model may have a modular architecture including one or more building units. In some examples, the machine learning model may have a multi-branch architecture. In some examples, the noise may be identified and removed from the image by an iterative process.

Abstract: This application describes a spatial envelope defining interaction of a robot arm and a fixture. The spatial envelope has a “keep-out” envelope defining a volume that the robot arm shall avoid entering, an input location defining a volume where the robot arm is to place a first object, an output location defining a volume from which the robot arm is to remove a second object, and a plurality of target items.

Abstract: A method includes generating a plurality of lidar inputs based on the lidar data, where each lidar input from among the plurality of lidar inputs comprises an image-based portion and a geometric-based portion, and where each lidar input from among the plurality of lidar inputs defines a position coordinate of the one or more objects. The method includes performing, for each lidar input from among the plurality of lidar inputs, a convolutional neural network (CNN) routine based on the image-based portion to generate one or more image-based outputs and assigning the plurality of lidar inputs to a plurality of echo groups based on the geometric-based portion. The method includes concatenating the one or more image-based outputs and the plurality of echo groups to generate a plurality of fused outputs and identifying the one or more objects based on the plurality of fused outputs.

Abstract: Described herein are computer-implemented systems, methods, and apparatuses for predicting classifications for image data, the system comprising a processor, memory and storage, and configured to: receive at least one image data sample; generate, using a deep neural network, for each image data sample, an activation layer comprising a set of activation vectors; perform activation vector correspondence for at least one of the activation vectors based on activation vector distance and conditioned by non-image data; and predict at least one classification for each image data sample based on the activation vector correspondence.

Abstract: Systems and techniques are provided for linking subjects in an area of real space with user accounts. The user accounts are linked with client applications executable on mobile computing devices. A plurality of cameras are disposed above the area. The cameras in the plurality of cameras produce respective sequences of images in corresponding fields of view in the real space. A processing system is coupled to the plurality of cameras. The processing system includes logic to determine locations of subjects represented in the images. The processing system further includes logic to match the identified subjects with user accounts by identifying locations of the mobile computing devices executing client applications in the area of real space and matching locations of the mobile computing devices with locations of the subjects.

Abstract: A system for updating a training dataset of an item identification model determines that an item is not included in a training dataset. In response to determining that the item is not included in the training dataset, the system obtains an identifier of the item. The system detects a triggering event at a platform, where the triggering event corresponds to a user placing the item on a platform. The system captures images of the item. The system extracts a set of features associated with the item from the images. The system associates the item to the identifier and the set of features. The system adds a new entry to the training dataset, where the new entry represents the item labeled with the identifier and the set of features.

Abstract: The present disclosure relates to an image processing device, an image processing method, and a program that can reduce the amount of processing required for a series of processing from detection to recognition of an object in a high-resolution image. An acquisition unit acquires, from a first resolution image, a second resolution image having a lower resolution than the first resolution image, a classification unit classifies an object included in the second resolution image, an identification unit identifies an object area corresponding to the object of a predetermined classification in the first resolution image, and a recognition unit performs recognition processing of the object on the object area identified in the first resolution image. The technology according to the present disclosure can be applied to a camera system of a remote control tower, for example.

Abstract: A method of using machine learning to output task-specific predictions may include receiving a digitized cytology image of a cytology sample and applying a machine learning model to isolate cells of the digitized cytology image. The machine learning model may include identifying a plurality of sub-portions of the digitized cytology image, identifying, for each sub-portion of the plurality of sub-portions, either background or cell, and determining cell sub-images of the digitized cytology image. Each cell sub-image may comprise a cell of the digitized cytology image, based on the identifying either background or cell. The method may further comprise determining a plurality of features based on the cell sub-images, each of the cell sub-images being associated with at least one of the plurality of features, determining an aggregated feature based on the plurality of features, and training a machine learning model to predict a target task based on the aggregated feature.

Abstract: Systems and methods for producing high-quality image segmentation masks by efficiently augmenting low-quality segmentation masks initially obtained by lightly trained neural networks are disclosed. Sequences of low-quality segmentation masks are transformed into high-quality segmentation masks by first identifying groups of pixels that are consistently segmented in at least three successive video frames. The pixels surrounding the consistently segmented pixels are then examined and assigned gradient alpha values, i.e., degrees of transparency. In one embodiment the alpha values are made proportional to the RGB distance of the pixel from the mean RGB value of the consistently segmented group of pixels.

Abstract: Disclosed is a method for inspecting a food product, the method includes: receiving image data representing the food product captured with an X-ray imaging unit; performing a texture analysis to image data for generating a first set of detections; performing a pattern analysis to at least part of the image data, the pattern analysis performed with a machine-learning component trained to identify objects with predefined pattern, for generating a second set of detections; generating an indication of an outcome of an inspection of the food product in accordance with a combination of the generated first set of detections and the second set of detections. Also disclosed is an apparatus and a computer program product.

Abstract: A device configured to receive a first point cloud data for a first item, to identify a first plurality of data points for the first object within the first point cloud data, and to extract the first plurality of data points from the first point cloud data. The device is further configured to receive a second point cloud data for the first item, to identify a second plurality of data points for the first object within the second point cloud data, and to extract a second plurality of data points from the second point cloud data. The device is further configured to merge the first plurality of data points and the second plurality of data points to generate combined point cloud data and to determine dimensions for the first object based on the combined point cloud data.

Abstract: A device configured to capture a first image of an item on a platform using a camera and to determine a first number of pixels in the first image that corresponds with the item. The device is further configured to capture a first depth image of an item on the platform using a three-dimensional (3D) sensor and to determine a second number of pixels within the first depth image that corresponds with the item. The device is further configured to determine that the difference between the first number of pixels in the first image and the second number of pixels in the first depth image is less than the difference threshold value, to extract the plurality of pixels corresponding with the item in the first image from the first image to generate a second image, and to output the second image.

Abstract: An object identification method is applied to a surveillance system. The surveillance system includes at least one surveillance apparatus. The object identification method includes acquiring a plurality of first feature vectors of a first moving object and a plurality of second feature vectors of at least one second moving object within a series of surveillance images from the surveillance apparatus, transforming the first feature vectors and the second feature vectors respectively into a first cluster distribution set and at least one second cluster distribution set, comparing similarity of the first cluster distribution set and the at least one second cluster distribution set, and setting a similarity ranking of the first cluster distribution set and the at least one second cluster distribution set according to a comparison result so as to determine whether the first moving object and the at least one second moving object are the same.

Abstract: A panorama rendering method, an electronic device and a storage medium are provided, and relates to the field of panoramic technology. The method includes: segmenting a panoramic picture to obtain segmented panoramic tile pictures; determining a target tile picture under a current screen and a target spherical patch in a panoramic picture sphere model, according to the panoramic picture sphere model and the segmented panoramic tile pictures; and pelfonning drawing on the target spherical patch by using the target tile picture, to obtain a target scene. In this way, time consumption of panorama loading can be reduced and loading performance can be improved.

Abstract: Technologies are provided herein for adaptively learning and matching modalities of face images. A face image is obtained and a biometric signature is generated of the face image. The biometric signature is compared with a stored biometric signature obtained from data storage. Differences are determined between the biometric signature and the stored biometric signature and compared with a proximity threshold. As a result of the comparison with the proximity threshold, the identity of the person in the face image may be determined or the face image may be stored in data storage in association with the identity of the person.

Abstract: Systems and methods include processors for receiving training data for a user activity; receiving bias criteria; determining a set of model parameters for a machine learning model including: (1) applying the machine learning model to the training data; (2) generating model prediction errors; (3) generating a data selection vector to identify non-outlier target variables based on the model prediction errors; (4) utilizing the data selection vector to generate a non-outlier data set; (5) determining updated model parameters based on the non-outlier data set; and (6) repeating steps (1)-(5) until a censoring performance termination criterion is satisfied; training classifier model parameters for an outlier classifier machine learning model; applying the outlier classifier machine learning model to activity-related data to determine non-outlier activity-related data; and applying the machine learning model to the non-outlier activity-related data to predict future activity-related attributes for the user activity

Abstract: A device configured to identify a first pixel location within a first plurality of pixels corresponding with an item in a first image and to apply a first homography to the first pixel location to determine a first (x,y) coordinate. The device is further configured to identify a second pixel location within a second plurality of pixels corresponding with the item in a second image and to apply a second homography to the second pixel location to determine a second (x,y) coordinate. The device is further configured to determine that the distance between the first (x,y) coordinate and the second (x,y) coordinate is less than or equal to the distance threshold value, to associate the first plurality of pixels and the second plurality of pixels with a cluster for the item, and to output the first plurality of pixels and the second plurality of pixels.

Abstract: Various examples are provided related to identification of anthropometric information for fitting of compression garments. In one example, a method of generating compression garment fit information includes acquiring images including a selected body part or body area in need of compression therapy; processing the acquired images along with a library of compression garment fit information; generating a 3D reconstruction of the selected body part/area; deriving anthropometric information for the selected body part/area from the 3D reconstruction; providing a compression value corresponding to a prescribed or intended amount of compression therapy applied to the selected body part/area; and generating compression garment fit information for the selected body part/area.

Abstract: Disclosed in the present disclosure is a method for processing remote sensing images by fusing elevation information.