Abstract: A method for retrieving heights of densely-covered forest canopies near power transmission lines includes: acquiring ICESat-2 LiDAR data, JL-1 image data, auxiliary data and three-dimensional information data about the power transmission lines within a target area; carrying out image preprocessing on a JL-1 image to generate a first image; screening the ICESat-2 LiDAR data according to a screening rule to obtain high-quality laser tree height data; by employing the first image, the high-quality laser tree height data and the auxiliary data, training a neural network model for retrieving a forest tree height according to an optical image and the auxiliary data; by employing the neural network model, generating a height distribution map of densely-covered trees in the target area; and calculating a height difference between a tree and a power transmission line nearest the tree to generate a hidden danger troubleshooting theme map.

Abstract: Current pretrained vision-language models for cross-modal retrieval tasks in English depend upon on the availability of many annotated image-caption datasets for pretraining to have English text. However, the texts are not necessarily in English. Although machine translation (MT) tools may be used to translate text to English, the performance largely relies on MT's quality and may suffer from high latency problems in real-world applications. Embodiments herein address these problems by learning cross-lingual cross-modal representations for matching images and their relevant captions in multiple languages. Embodiments seamlessly combine cross-lingual pretraining objectives and cross-modal pretraining objectives in a unified framework to learn image and text in a joint embedding space from available English image-caption data, monolingual corpus, and parallel corpus. Embodiments are shown to achieve state-of-the-art performance in retrieval tasks on multimodal multilingual image caption datasets.

Abstract: A method for automatically identifying faulty signatures for autonomous driving applications, the method includes receiving, by a processing circuit, a signature; matching the signature to a group of first signatures that are untagged and are randomly obtained; identifying, based on the matching, first top matching signatures; matching the signature to a group of second signatures that are untagged and are correctly or erroneously indicative of a detection of reference elements; identifying, based on the matching of the signature to the second group of second signatures, second top matching signatures; determining an overlap between the first top matching signatures and the second top matching signatures; and determining whether the signature is faulty or faultless based on the overlap.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a IT machine learning model. One of the methods includes receiving training data comprising a plurality of unlabeled training inputs and a plurality of labeled training inputs; generating augmented training data, comprising generating, for each of the plurality of unlabeled training inputs, a respective augmented training input by applying a data augmentation technique to the unlabeled training input; and training the machine learning model on the augmented training data. In particular, but not exclusively, the model may be trained for perceptual tasks (e.g. tasks relating to vision or speech).

Abstract: Devices, systems, and methods are provided for brand matching using multi-layer machine learning. A method may include generating, based on a first embedding vector and a second embedding vector as inputs to a twin neural network, a third embedding vector and a fourth embedding vector; generating, based on the first embedding vector and the second embedding vector as inputs to a difference neural network, a difference vector indicative of a difference between the first embedding vector and the second embedding vector; generating a concatenated vector by concatenating the third embedding vector with the fourth embedding vector and the difference vector; generating, based on the concatenated vector as an input to a feedforward neural network (FFN), a score between zero and one, the score indicative of a relationship between a first entity and a second entity.

Abstract: A multiple dwelling house interphone system includes a collective entrance machine, a dwelling room master device, and a management master device. The collective entrance machine is for calling a visit destination by a visitor. The dwelling room master device is for responding to the call by a dweller. The management master device is for speaking with the dweller by a manager. The management master device includes a speech permissibility storage section and a speech control section. The speech permissibility storage section registers a permissibility of calling and speaking to the dwelling room master device for each dwelling unit. The speech control section controls the calling and speaking. The speech control section does not perform the calling control when a calling operation by the management master device is performed towards the dwelling room master device of a dwelling unit registered as “unpermitted” in the speech permissibility storage section.

Abstract: Disclosed are an astronomical telescope stand, an auxiliary calibration method for the astronomical telescope and an astronomical telescope system. The astronomical telescope includes at least two rotating shafts perpendicular to each other, each rotating shaft is provided with an encoder; an electronic camera shooting in the same direction as the telescope to be mounted; a display screen; a processor that connects with the encoder, the electronic camera and the display screen; the processor includes an image information parsing module, an electronic image generation module and a coordinate correction module. The telescope is mounted on the stand, and the electronic camera shoots the same image as the image observed by the telescope. The image information parsing module solves the image to obtain the center coordinate and the field size of the image. The electronic image generation module generates the simulated image according to the center coordinate and the field size.

Abstract: An information processing apparatus in the present invention includes: an acquisition unit that, based on registration information including a plurality of registrants and registration time associated with each of the plurality of registrants and on the current time, acquires a first biometrics information group including biometrics information on a candidate of a matching process from a registered biometrics information group including biometrics information on the plurality of registrants; and a matching unit that matches biometrics information on a person detected from an image captured in a matching area against biometrics information included in the first biometrics information group.

Abstract: An image processing apparatus 1 includes a pre-processor 10 and a main processor 30. The pre-processor 10 generates pre-processing data by performing a pre-processing of Bayer data outputted from an imaging device 91, and outputs the generated pre-processing data. The main processor 30 performs an image processing based on the pre-processing data.

Abstract: A computer-implemented method includes receiving, by a computing device, a particular textual description of a scene. The method also includes applying a neural network for text-to-image generation to generate an output image rendition of the scene, the neural network having been trained to cause two image renditions associated with a same textual description to attract each other and two image renditions associated with different textual descriptions to repel each other based on mutual information between a plurality of corresponding pairs, wherein the plurality of corresponding pairs comprise an image-to-image pair and a text-to-image pair. The method further includes predicting the output image rendition of the scene.

Abstract: An event detection system to operations that include: detecting an event based on at least a portion of the sensor data, the event comprising a plurality of event attributes; determining one or more event attributes from the plurality of event attributes of the event transgress a threshold value; and causing display of a notification that includes a presentation of at least the portion of the sensor data in response to the determining that the one or more event attributes from among the plurality of event attributes of the event transgress the threshold value.

Abstract: In image processing system comprises an input configured to receive an image signal, the image signal including a plurality of frames of image data; and a processor configured to automatically determine an image classification based on at least one frame of the plurality of frames, and dynamically generate a mapping metadata based on the image classification. The processor includes determination circuitry configured to determine a content type for the image signal; segmentation circuitry configured to segment the image data into a plurality of feature item regions, based on the content type; extraction circuitry configured to extract at least one image aspect value for respective ones of the plurality of feature item regions.

Abstract: The present disclosure involves systems, software, and computer implemented methods for transaction auditing. One example method includes training at least one machine learning model to determine features that can be used to determine whether an image is an authentic image of a document or an automatically generated document image, using a training set of authentic images and a training set of automatically generated document images. A request to classify an image as either an authentic image of a document or an automatically generated document image is received. The machine learning model(s) are used to classify the image as either an authentic image of a document or an automatically generated document image, based on features included in the image that are identified by the machine learning model(s). A classification of the image is provided. The machine learning model(s) are updated based on the image and the classification of the image.

Abstract: The description relates the feature matching. Our approach establishes pointwise correspondences between challenging image pairs. It takes off-the-shelf local features as input and uses an attentional graph neural network to solve an assignment optimization problem. The deep middle-end matcher acts as a middle-end and handles partial point visibility and occlusion elegantly, producing a partial assignment matrix.

Abstract: A printer-verifier is provided for printing and verifying a printed indicium. Printer-verifier includes imaging module for capturing image of printed indicium on print media, memory communicatively coupled to imaging module and configured to store print quality verification program, and processor communicatively coupled to memory. Processor is configured by print quality verification program to evaluate print quality of printed indicium, determine if print quality meets print quality standard, reject printed indicium on print media as rejected printed media when print quality does not meet print quality standard, and one of output location-specific feedback about rejected printed media or send a command signal to cutting device to destroy rejected printed media. Location-specific feedback is in same language as language associated with printer-verifier. A printer and methods for verifying the printed indicium are also provided.

Abstract: Example implementations described herein involve an attribute of an inflow source that is assigned to each trajectory in the camera view, which is used for crowd analysis. The inflow source is estimated using techniques of tracking across cameras. If the inflow source is different even if the flow trajectories are the same direction, the trajectories are visualized in different styles. By using the attribute of the inflow source, it is possible to detect changes and anomalies in the crowd that were difficult to detect using the prior art.

Abstract: According to an exemplary embodiment of the present disclosure, a method of detecting an object is disclosed. The method of detecting an object includes computing an image including an object by using an object detection model including a local block and a non-local block, in which the local block computes a relationship between adjacent pixels included in a feature map, and the non-local block computes a relationship between non-adjacent pixels included in the feature map.

Abstract: One variation of a method for monitoring manufacture of assembly units includes: receiving selection of a target location hypothesized by a user to contain an origin of a defect in assembly units of an assembly type; accessing a feature map linking non-visual manufacturing features to physical locations within the assembly type; for each assembly unit, accessing an inspection image of the assembly unit recorded by an optical inspection station during production of the assembly unit, projecting the target location onto the inspection image, detecting visual features proximal the target location within the inspection image, and aggregating non-visual manufacturing features associated with locations proximal the target location and representing manufacturing inputs into the assembly unit based on the feature map; and calculating correlations between visual and non-visual manufacturing features associated with locations proximal the target location and the defect for the set of assembly units.

Abstract: A method and a system for producing a story video are provided. A method for producing a story video, according to one embodiment, can produce a specific story video by determining a theme of a story that is suitable for collected videos and selecting and arranging an appropriate video for each frame of a template associated with the theme.

Abstract: Disclosed is a multi-mode communication notification system and method for improving communication between an inmate and a third party by enabling a third party to send a web-initiated contact request for a specific inmate and further allowing the inmate to check his/her contact requests or be automatically notified of contact requests. Additionally, the system electronically notifies the third party when the specific inmate has been notified of the contact request. Further, the system notifies the third party if they have missed a call from the specific inmate.