Abstract: Efficient location and identification of documents in images. In an embodiment, at least one quadrangle is extracted from an image based on line(s) extracted from the image. Parameter(s) are determined from the quadrangle(s), and keypoints are extracted from the image based on the parameter(s). Input descriptors are calculated for the keypoints and used to match the keypoints to reference keypoints, to identify classification candidate(s) that represent a template image of a type of document. The type of document and distortion parameter(s) are determined based on the classification candidate(s).

Abstract: A method includes receiving selection of a target within an image captured by an image sensor of a payload and displayed on a user interface of the payload, detecting a deviation of the target from an expected target state within the image, generating, based at least partly on the deviation, a payload control signal including a first angular velocity for rotating the payload about an axis of the carrier to reduce the deviation about the axis in a subsequent image, and generating a base support control signal including a second angular velocity for rotating the payload with respect to the axis. When the first and second angular velocities are received, the carrier is controlled to rotate the payload at a third angular velocity about the axis. The third angular velocity is the first angular velocity, the second angular velocity, or a combination of both.

Abstract: Methods, systems, articles of manufacture, and apparatus to extract shape features based on a structural angle template are disclosed. An example apparatus includes a template generator to generate a template based on an input image and calculate a template value based on values in the template; a bit slicer to calculate an OR bit slice and an AND bit slice based on the input image, combine the OR bit slice with the AND bit slice to generate a fused image, group a plurality of pixels of the fused image to generate a pixel window, each pixel of the pixel window including a pixel value, and calculate a window value based on the pixel values of the pixel window; and a comparator to compare the template value with the window value and store the pixel window in response to determining the window value satisfies a similarity threshold with the template value.

Abstract: An angled optical pattern is decoded. To decode an optical pattern imaged at an angle, an area of interest of an image is received. A start line and an end line of the optical pattern are estimated. Corners of the optical pattern are localized. A homography is calculated based on the corners. And a scanline of the optical pattern is rectified based on the homography.

Abstract: An object is to provide a polygon detection device, a polygon detection method, and a polygon detection program to accurately detect a polygon resembling a reference polygon from an image. The polygon detection device acquires a ratio among lengths of sides of a reference polygon included in an appearance of a predetermined object. The polygon detection device acquires a photographic image of the predetermined object. The polygon detection device detects line segments from the acquired photographic image. The polygon detection device forms at least one polygon based on the detected line segments. The polygon detection device identifies, from the formed polygon, a polygon corresponding to the reference polygon based on a degree of similarity between a ratio among lengths of sides of the formed polygon and the acquired ratio among the lengths of sides of the reference polygon, among from the formed polygon.

Abstract: A domain alignment technique for cross-domain object detection tasks is introduced. During a preliminary pretraining phase, an object detection model is pretrained to detect objects in images associated with a source domain using a source dataset of images associated with the source domain. After completing the pretraining phase, a domain adaptation phase is performed using the source dataset and a target dataset to adapt the pretrained object detection model to detect objects in images associated with the target domain. The domain adaptation phase may involve the use of various domain alignment modules that, for example, perform multi-scale pixel/path alignment based on input feature maps or perform instance-level alignment based on input region proposals.

Abstract: A facial recognition method using online sparse learning includes initializing target position and scale, extracting positive and negative samples, and extracting high-dimensional Haar-like features. A sparse coding function can be used to determine sparse Haar-like features and form a sparse feature matrix, and the sparse feature matrix in turn is used to classify targets.

Abstract: Techniques for providing improved optical character recognition (OCR) for receipts are discussed herein. Some embodiments may provide for a system including one or more servers configured to perform receipt image cleanup, logo identification, and text extraction. The image cleanup may include transforming image data of the receipt by using image parameters values that optimize the logo identification, and performing logo identification using a comparison of the image data with training logos associated with merchants. When a merchant is identified, a second image clean up may be performed by using image parameter values optimized for text extraction. A receipt structure may be used to categorize the extracted text. Improved OCR accuracy is also achieved by applying on format rules of the receipt structure to the extracted text.

Abstract: Methods and apparatus to determine the dimensions of a region of interest of a target object and a class of the target object from an image using target object landmarks are disclosed herein. An example method includes identifying a landmark of a target object in an image based on a match between the landmark and a template landmark; classifying a target object based on the identified landmark; projecting dimensions of the template landmark based on a location of the landmark in the image; and determining a region of interest based on the projected dimensions, the region of interest corresponding to text printed on the target object.

Abstract: Disclosed are a method and apparatus for presenting material, and a storage medium. The method includes acquiring at least two key points from a position of a presentation part of an object in an image; determining a preselected target point based on positions of the at least two key points; determining a target point of the image based on the preselected target point and target points of N continuous frames before the image, and presenting the material based on the target point.

Abstract: A method for real-time video processing for changing features of an object in a video, the method comprises: providing an object in the video, the object being at least partially and at least occasionally presented in frames of the video; detecting the object in the video; generating a list of at least one element of the object, the list being based on the object's features to be changed according to a request for modification; detecting the at least one element of the object in the video; tracking the at least one element of the object in the video; and transforming the frames of the video such that the at least one element of the object is modified according to the request for modification.

Abstract: Disclosed is a system, method, and devices as system elements to recognize an object by an object recognizing system including an imaging device and a moving assembly to move the imaging device around the object, to form a certified visual model of the object to be recognized. Especially the disclosure relates to gemstone imaging by an imaging method including photographing a target, in an illumination, by a camera, to obtain at least one image of the targeted object to be recognized.

Abstract: The present disclosure relates to information processing apparatus and method that makes it possible to suppress a reduction in encoding efficiency. Information relating to quantization of a three-dimensional position of an encoding target is generated. For example, the information relating to the quantization includes information relating to a coordinate system to be subjected to the quantization, information relating to a bounding box for normalization of position information of the encoding target, or information relating to a voxel for quantization of position information of the encoding target. In addition, three-dimensional information of the encoding target is restored from a signal string on the basis of the information relating to the quantization of the three-dimensional position of the encoding target.

Abstract: A method and an apparatus for asynchronous data fusion, a storage medium and an electronic device are provided. The method includes: obtaining current frame LiDAR data, and determining current frame LiDAR three-dimensional embeddings; determining a previous frame fused hidden state, and performing a temporal fusion process based on the previous frame fused hidden state and the current frame LiDAR three-dimensional embeddings to generate a current frame temporary hidden state and a current frame output result; and obtaining current frame camera data, determining current frame camera three-dimensional embeddings, and generating a current frame fused hidden state based on the current frame camera three-dimensional embeddings and the current frame temporary hidden state. Asynchronous fusion is performed on the current frame LiDAR data and previous frame camera data, which leads to a low processing latency.

Abstract: There is a need for more effective and efficient document processing solution. Accordingly, various embodiments of the present invention introduce various document processing optimization solutions. In one example, a method includes identifying a plurality of input pages each associated with a related input document of a plurality of input documents; for each input page of the plurality of input pages, generating a segmented page; processing each segmented page using a trained encoder model to generate a fixed-dimensional representation of the input page; determining, based at least in part on each fixed-dimensional representation, a plurality of document clusters; determining a plurality of processing groups, where each processing group is associated with one or more related document clusters of the plurality of document clusters; and performing the document processing optimization based at least in part on the plurality of processing groups.

Abstract: In at least one embodiment, a computer-program product embodied in a non-transitory computer readable medium that is programmed to detect a performance threshold for one or more cameras is provided. The computer-program product includes instructions to capture a plurality of images from one or more cameras to compare an object within each captured image to a predetermined object to determine whether the object has been correctly identified and instructions to extract the object from each captured image. The computer-program product includes instructions to apply at least one gradient to each extracted object to generate a plurality of gradient images. The computer-program product includes instructions to compare the extracted object to the predetermined object and to determine whether the extracted object that is modified by the at least one gradient has been correctly identified. The computer-program product includes instructions to establish a performance threshold for the one or more cameras.

Abstract: Some implementations relate to determining whether glare is present in captured image(s) of an object (e.g., a photo) and/or to determining one or more attributes of any present glare. Some of those implementations further relate to adapting one or more parameters for a glare removal process based on whether the glare is determined to be present and/or based on one or more of the determined attributes of any glare determined to be present. Some additional and/or alternative implementations disclosed herein relate to correcting color of a flash image of an object (e.g., a photo). The flash image is based on one or more images captured by a camera of a client device with a flash component of the client device activated. In various implementations, correcting the color of the flash image is based on a determined color space of an ambient image of the object.

Abstract: The present disclosure describes techniques for generating a mask frame data segment corresponding to a video frame. The disclosed techniques include obtaining a frame of a video; identifying a main area of the frame using an image segmentation algorithm; and generating a mask frame data segment corresponding to the frame based on the main area of the frame, wherein the generating a mask frame data segment corresponding to the frame based on the main area of the frame further comprises generating the mask frame data segment based on a timestamp of the frame in the video, a width and a height of the main area of the frame.

Abstract: Method for real time video processing for retouching an object in a video, comprising: providing an object in the video steam, the object being at least partially and at least occasionally presented in frames of the video; setting a degree of retouching; generating a list of at least one element of the object selected based on a request of retouching and the degree of retouching; detecting the at least one element of the object in the video and parameters of the at least one element; calculating new parameters of the at least one element according to the degree of retouching; detecting characteristic points for each of the at least one element of the object; generating a mesh based on the characteristic points for each of the at least one element of the object; tracking the at least one element of the object in the video, wherein the tracking comprises aligning the mesh for each of the at least one element with a position of the corresponding each of the at least one element; and transforming the frames of th

Abstract: Systems, methods, tangible non-transitory computer-readable media, and devices associated with object localization and generation of compressed feature representations are provided. For example, a computing system can access training data including a target feature representation and a source feature representation. An attention feature representation can be generated based on the target feature representation and a machine-learned attention model. An attended target feature representation can be generated based on masking the target feature representation with the attention feature representation. A matching score for the source feature representation and the target feature representation can be determined. A loss associated with the matching score and a ground-truth matching score for the source feature representation and the target feature representation can be determined. Furthermore, parameters of the machine-learned attention model can be adjusted based on the loss.