Abstract: Provided is a virtual object processing method. The virtual object processing method includes: detecting a spatial plane in a scene where a first device is located; detecting real objects in the scene to determine a plurality of real object position boxes; determining, based on a matching relation between the plurality of real object position boxes and the spatial plane in the scene, a candidate position box set from the plurality of real object position boxes; determining, in response to a virtual object configuration operation for a target position box in the candidate position box set, position information of the virtual object in the target position box; transmitting position on the virtual object and the position information of the virtual object in the target position box to a second device for displaying the virtual object on the second device.

Abstract: A mobile device is fitted with a camera and an extended reality (XR) software application program executing on a processor within an XR system. Via the XR software application program, various techniques are performed for manipulating virtual objects in an XR environment. In a first technique, the XR software application program facilitates the movement of a virtual object from a first location to a second location. In a second technique, the XR software application program facilitates the rotation of a virtual object. In a third technique, the XR software application program facilitates the scaling of a virtual object along one or more axes.

Abstract: An image processing device according to one embodiment estimates optical flow information, pixel by pixel, on the basis of a reference image and input images of consecutive frames, and estimates a term corresponding to temporal consistency between the frames of the input images. The image processing device determines a mesh on the basis of the term corresponding to temporal consistency and the optical flow information, and transforms the reference image on the basis of the mesh. The image processing device preforms image blending on the basis of the input image, the transformed reference image, and mask data.

Abstract: Moving images of a space, which includes objects 34 and 35 of a display target, as viewed from reference points are created in advance as reference images, and they are combined in response to actual positions of the points of view to draw a moving image. When the object 35 is displaced as indicated by an arrow mark in the space, reference points of view 30a to 30e are fixed as depicted in (a). Alternatively, the reference points of view are displaced in response to the displacement like reference points of view 36a to 36e in (b). Then, the moving images from the reference points of view are generated as the reference images.

Abstract: An image correction method includes: acquiring band images obtained by imaging a subject, and a high-resolution image having a resolution higher than that of the band images; acquiring a position difference between the object band image and the reference band image among the band images; by using a pixel of the object band image as an object pixel, for each object pixel, determining a pixel value of each sub-region obtained by dividing the imaging region of the object pixel into a plurality of regions, based on the pixel value of the object pixel and a relationship between pixel values of the pixels of the high-resolution image corresponding to the object pixel; and creating a corrected band image that holds a pixel value of light on the object band image at the pixel position of the reference band image, from the determined pixel value of each sub-region and the position difference.

Abstract: Images of an undercarriage of a vehicle may be captured via one or more cameras. A point cloud may be determined based on the images. The point cloud may includes points positioned in a virtual three-dimensional space. A stitched image may be determined based on the point cloud by projecting the point cloud onto a virtual camera view. The stitched image may be stored on a storage device.

Abstract: Provided are an image processing method and apparatus, a device, and a storage medium, relating to the technical field of image processing, in particular to the artificial intelligence fields such as computer vision and deep learning. The specific implementation scheme is as follows: inputting a to-be-processed image into an encoding network to obtain a basic image feature, wherein the encoding network includes at least two cascaded overlapping encoding sub-networks which perform encoding and fusion processing on input data at at least two resolutions; and inputting the basic image feature into a decoding network to obtain a target image feature for pixel point classification, wherein the decoding network includes at least one cascaded overlapping decoding sub-network to perform decoding and fusion processing on input data at at least two resolutions respectively.

Abstract: Methods and systems are disclosed for quantizing images using machine-learning. A plurality of input images are received from a sensor (e.g., a camera), wherein each input image includes a plurality of pixels. Utilizing an image-to-image machine-learning model, each pixel is assigned a new pixel color. Utilizing a mixer machine-learning model, each new pixel color is converted to one of a fixed number of colors to produce a plurality of quantized images, with each quantized image corresponding to one of the input images. A loss function is determined based on an alignment of each input image with its corresponding quantized image via a pre-trained reference machine-learning model. One or more parameters of the image-to-image machine-learning model and the mixer model are updated based on the loss function. The process repeats, with each iteration updating the parameters of the image-to-image machine-learning model and the mixer model, until convergence, resulting in trained models.

Abstract: The disclosure provides an image reconstruction method for an edge device, an electronic device and a storage medium. The image reconstruction method includes: extracting low-level features from an input image of a first scale to generate first feature maps, the first feature maps having a second scale greater than the first scale as compared with the input image; extracting low-level features from the input image to generate second feature maps, the second feature maps having the second scale; generating mask maps based on the second feature maps; generating intermediate feature maps based on the mask maps and the first feature maps, the intermediate feature maps having the second scale; synthesizing a reconstructed image having the second scale based on the intermediate feature maps. This method facilitates to achieve a better image super-resolution reconstruction effect with lower resource consumption.

Abstract: Methods and systems include neural network-based image processing and blending circuitry to blend an output of the neural network to compensate for potential artifacts from the neural network-based image processing. The neural network(s) apply image processing to image data using one or more neural networks as processed data. Enhance circuitry enhances the image data in a scaling circuitry to generate enhanced data. Blending circuitry receives the processed image data and the enhanced data along with an image plane of the processed data. The blending circuitry also determines whether the image processing using the one or more neural networks has applied a change to the image data greater than a threshold amount. The blending circuitry then, based at least in part in response to the change being greater than the threshold amount and/or edge information of the image data, blends the processed data with the enhanced data.

Abstract: An electronic device is provided. The electronic device includes an input interface and a processor that performs noise removal processing on an image input through the input interface, obtains first information on a first texture block among a plurality of first pixel blocks included in the input image, obtains second information on a second texture block among a plurality of second pixel blocks included in the noise-removed image, obtains third information on maximum energy amount among energy amount of each of the plurality of second pixel blocks, and identifies whether the input image is an upscaled image based on the first, second and third information.

Abstract: Augmented Denoising Diffusion Implicit Models (“DDIMs”) using a latent trajectory optimization process can be used for image generation and manipulation using text input and one or more source images to create an output image. Noise bias and textual bias inherent in the model representing the image and text input is corrected by correcting trajectories previously determined by the model at each step of a diffusion inversion process by iterating multiple starts the trajectories to find determine augmented trajectories that minimizes loss at each step. The trajectories can be used to determine an augmented noise vector, enabling use of an augmented DDIM and resulting in more accurate, stable, and responsive text-based image manipulation.

Abstract: Methods, systems, and apparatus for removing precipitation from video are disclosed. A method includes generating, from a first set of images of a scene from a camera, a segmented background image model of the scene; obtaining a second set of images from the camera; identifying, in an image of the second set of images, a plurality of edges, determining that a first edge of the plurality of edges satisfies criteria for representing precipitation based at least in part on determining that the first edge (i) does not correspond to the background image model of the scene and (ii) extends into two or more contiguous segments of the scene; in response, classifying each of the contiguous segments as a precipitation segment; generating pixel data for each of the precipitation segments; and applying the pixel data to each precipitation segment in the image.

Abstract: A control device for a mobile object including one or more imaging devices, the control device includes an image acquisition unit configured to acquire an image of an external environment of the mobile object from the one or more imaging devices, a correction unit configured to perform distortion reduction processing for reducing distortion of an image for each of one or more regions included in an image acquired from the one or more imaging devices, and a recognition unit configured to recognize the external environment of the mobile object based on an image on which the distortion reduction processing has been performed. The correction unit is configured to determine the one or more regions to be a target of the distortion reduction processing in accordance with a predetermined rule according to a moving scene of the mobile object.

Abstract: An image processing apparatus comprises: an acquisition unit that acquires a first image obtained through shooting and distance information of the first image; a detection unit that detects a main subject from the first image; an extraction unit that extracts another subject from the first image based on the distance information of the main subject; a setting unit that sets parameters of one or more virtual light sources that emit virtual light to the main subject and the extracted other subject; and a processing unit that generates from the first image a second image in which the main subject and the other subject are illuminated with the virtual light using the parameters set by the setting unit.

Abstract: This disclosure describes one or more embodiments of systems, non-transitory computer-readable media, and methods that can learn or identify a learned-initialization-latent vector for an initialization digital image and reconstruct a target digital image using an image-generating-neural network based on a modified version of the learned-initialization-latent vector. For example, the disclosed systems learn a learned-initialization-latent vector from an initialization image utilizing a high number (e.g., thousands) of learning iterations on an image-generating-neural network (e.g., a GAN). Then, the disclosed systems can modify the learned-initialization-latent vector (of the initialization image) to generate modified or reconstructed versions of target images using the image-generating-neural network.

Abstract: A method of recognizing a specific image includes detecting image features from the specific image and storing in a memory each of detected relevant image features, detecting relevant image features from an input image and selecting from the memory a portion corresponding to each of the detected relevant image features, generating a restoration image using portions selected in the selecting, and determining whether the generated restoration image matches with the input image by a matching process and recognizing that, when it is determined that the restoration image matches with the input image, the input image is the specific image.

Abstract: A single-shot differential phase contrast quantitative phase imaging method based on color multiplexing illumination. A color multiplexing illumination solution is used to realize single-shot differential phase contrast quantitative phase imaging. In the single-shot color multiplexing illumination solution, three illumination wavelengths of red, green, and blue are used to simultaneously illuminate a sample, and the information of the sample in multiple directions is converted into intensity information on different channels of a color image. By performing channel separation on this color image, the information about the sample at different spatial frequencies can be obtained.

Abstract: An image processing device receives a caption for an image before, during, or after capture of the image by an image capture device. The image processing device generates image processing settings based on the caption, for instance based on a mood indicated in the caption or an object identified in the caption. If the caption is received before image capture, the image processing settings may include image capture settings that the image capture device may use to alter exposure or focus during image capture. Once the image is captured, the image processing device may process the image based on the image processing settings, for instance by applying filters or adjusting gain, brightness, contrast, saturation, or colors. For instance, brightness and saturation may be altered if the caption indicates a happy or sad mood, and focus may be altered to focus on an object identified in a caption.

Abstract: A defect detection apparatus includes a defect detection unit and a calculation unit. The defect detection unit detects a defect that appears on an external appearance of a structure through imaging processing from external appearance image data that is generated by imaging the external appearance of the structure with a surveying instrument. The calculation unit calculates defect data relating to the defect detected by the defect detection unit by using coordinate data that is correlated with the external appearance image data.

Abstract: The present invention is directed to a method for quantitative and qualitative evaluation of laser printed samples, the method comprising the following steps: a) providing (S1) a laser printed sample, b) capturing (S2) a digital raster image of the laser printed part of the laser printed sample and thereon providing digital image information that constitutes the digital raster image, c) identifying (S3) at least one distinct part within the digital image information, d) obtaining at least one image histogram for the identified at least one distinct part of the digital image information, e) fitting (S4) a probability density function on the obtained at least one image histogram, and f) determining (S5) at least one parameter of the probability density function.

Abstract: Provided is a technique that allows an operator or the like to easily perceive a tendency of defects which are present in molds or patterns. A display device displays a result of inspection of a plurality of molds or a plurality of patterns which are inspection target objects. The display device classifies the inspection target objects into a plurality of groups, generates, for each of the plurality of groups, a heat map image from a result of defect inspection of one or more of inspection target objects which are classified into the each of the plurality of groups, the heat map image representing a spatial distribution of defect occurrence frequencies; and causes a display to display one or more of the heat map images generated, the one or more heat map images corresponding to one or more groups specified by a user from among the plurality of groups.

Abstract: A smart, human-centered technique that uses artificial intelligence and mixed reality to accelerate essential tasks of the inspectors such as defect measurement, condition assessment and data processing. For example, a bridge inspector can analyze some remote cracks located on a concrete pier, estimate their dimensional properties and perform condition assessment in real-time. The inspector can intervene in any step of the analysis/assessment and correct the operations of the artificial intelligence. Thereby, the inspector and the artificial intelligence will collaborate/communicate for improved visual inspection. This collective intelligence framework can be integrated in a mixed reality supported see-through headset or a hand-held device with the availability of sufficient hardware and sensors. Consequently, the methods reduce the inspection time and associated labor costs while ensuring reliable and objective infrastructure evaluation.

Abstract: Disclosed are a method for evaluating a geometric form of a honeycomb product and a system for detecting and evaluating a geometric form of a honeycomb product. The method for evaluating the geometric form includes: acquiring a top-surface image and a side-surface image of the honeycomb product; acquiring vertex coordinates by extracting vertices from the top-surface image; acquiring serial numbers of six vertices of each cell by reconstructing the cells of the top-surface image; computing deviation angles of six interior angles of each cell based on the serial numbers and vertex coordinates of the six vertices of each cell; extracting a top-surface side boundary and a side-surface side boundary of the honeycomb product from the top-surface image and the side-surface image; computing a maximum top-surface deflection and a maximum side-surface deflection of the honeycomb product based on the top-surface side boundary and the side-surface side boundary.

Abstract: An installation, for use with a computer, for inspecting pinion gear teeth, the installation comprising: an enclosure, the enclosure including a back, a top, a bottom, sides and a front, which includes an opening, to define an interior; a transparent window, the transparent window separating at least a part of the interior from an ambient environment; at least one door operatively connected to the enclosure and retractably separating the transparent barrier from the ambient environment; a programmable logic controller which is housed in the enclosure behind the transparent window; a thermal imager which is housed in the enclosure behind the transparent window, is directed to the transparent window and is in electronic communication with the programmable logic controller; a visible light camera which is housed in the enclosure behind the transparent window, is directed to the transparent window and is in electronic communication with the programmable logic controller; and an air blade blower which is attached

Abstract: The present disclosure includes systems, devices, and methods for identifying, detecting, and/or tracking rail features. In some aspects, a system includes a camera and a computer having at least one memory, at least one processor configured to receive a plurality of images from the camera, and for each of the images: assigning a location identifier and identifying one or more rail features that correspond to one of a plurality of predetermined rail features. In some systems, the at least one processor is configured to determine a location of each of the one or more identified rail features.

Abstract: Provided is a determination method capable of non-destructively and simply determining a state of a sphere that is an aggregate of a plurality of cells. A phase difference image of a sphere that is an aggregate of a plurality of cells is generated from a hologram obtained by imaging the sphere, and a state of the sphere is determined on the basis of the phase difference image and a shape index value corresponding to a shape of the sphere.

Abstract: Methods and apparatus for computer-aided prostate condition diagnosis are disclosed. An example computer-aided prostate condition diagnosis apparatus includes memory to store instructions and a processor. The example processor can detect a lesion from an image of a prostate gland and generate a mapping of the lesion from the image to a sector map, the generating the mapping of the lesion comprising identifying a depth region of the lesion, wherein the depth region indicates a location of the lesion along a depth axis. The processor can also provide the sector map comprising a representation of the lesion within the prostate gland mapped from the image to the sector map.

Abstract: A Method comprising the steps of, acquiring an input about the target object class to be searched within the image; associating a set of morphological and chromatic parameters to a pre-set initialization value range selected from a library on the basis of the input; processing the tissue image on the basis of the selected parameters via a first segmentation (based on object based thresholding, morphological filtering, active contouring) algorithm to identify objects within the desired target object class; providing a result set comprising the objects complying with parameters within the pre-set initialization value range; checking whether a count of objects within the result set is below a predefined threshold and either entering an initial loop if the count is below the predefined threshold, or entering an adaptation branch.

Abstract: Systems and methods described herein relate, among other things, to unmixing more than three stains, while preserving the biological constraints of the biomarkers. Unlimited numbers of markers may be unmixed from a limited-channel image, such as an RGB image, without adding any mathematical complicity to the model. Known co-localization information of different biomarkers within the same tissue section enables defining fixed upper bounds for the number of stains at one pixel. A group sparsity model may be leveraged to explicitly model the fractions of stain contributions from the co-localized biomarkers into one group to yield a least squares solution within the group. A sparse solution may be obtained among the groups to ensure that only a small number of groups with a total number of stains being less than the upper bound are activated.

Abstract: A computer apparatus and method for identifying and visualizing tumors in a histological image and measuring a tumor margin are provided. A CNN is used to classify pixels in the image according to whether they are determined to relate to non-tumorous tissue, or one or more classes for tumorous tissue. Segmentation is carried out based on the CNN results to generate a mask that marks areas occupied by individual tumors. Summary statistics for each tumor are computed and supplied to a filter which edits the segmentation mask by filtering out tumors deemed to be insignificant. Optionally, the tumors that pass the filter may be ranked according to the summary statistics, for example in order of clinical relevance or by a sensible order of review for a pathologist. A visualization application can then display the histological image having regard to the segmentation mask, summary statistics and/or ranking. Tumor masses extracted by resection are painted with an ink to highlight its surface region.

Abstract: Systems and methods for predicting a patient response to various agents and/or combinations of agents using ex vivo dosing and imaging are disclosed. In one example, a method of determining treatment efficacy includes analyzing a solid cell culture over time, e.g., first and second responses to a solid cell culture to respective treatments may be compared to determine a treatment efficacy of each treatment. Systems and methods for applying the treatments to the cell culture and analyzing the cell culture and efficacy are disclosed.

Abstract: The present embodiments relate generally to a probe, system, and method for generating a predictive model for moving the probe. The probe can include a movable element that move according to the suggestion of the predictive model. The system can include the probe, a user device, an administrator processor, and a server. The predictive model calculates an image score based on the quality, then the processor can move the movable element based on the score.

Abstract: A finding classification unit classifies each pixel of a first medical image into at least one finding. A feature amount calculation unit calculates a first feature amount for each finding. A weighting coefficient setting unit sets a weighting coefficient indicating a degree of weighting, which varies depending on a size of each finding, for each finding. A similarity derivation unit performs a weighting operation for the first feature amount for each finding calculated in the first medical image and a second feature amount for each finding calculated in advance in a second medical image on the basis of the weighting coefficient to derive a similarity between the first medical image and the second medical image.

Abstract: A hyperspectral facial analysis system and method for personalized health scoring to assess the risk that a person has a disease. Embodiments capture images in multiple spectral bands, such as visible, infrared, and ultraviolet, and analyze these images to generate multiple health metrics, such as pallor, temperature, sweat, and chromophores. These metrics may be combined into an overall health score that may be used for screening. Image analysis may focus on the area under the eyes, where skin is thinnest. Images may be compared to a reference population to identify anomalous values, so that health scoring automatically adjusts for local conditions. Pallor may be calculated based on hue and saturation of visible light images. Temperature may be calculated based on infrared image intensity. Sweat may be calculated using cross-polarized images to identify specular highlights. Chromophores may be calculated by comparing frequency domain ultraviolet images to those of the reference population.

Abstract: First and second images are displayed of anatomical structure segments with an attached fixator. Indications may be received of first image hinge locations of a plurality of hinges of the fixator in the first image. Projected second image hinge locations may be determined based at least in part on the first image hinge locations. Hinge candidates may be detected in the second image having shapes associated with the plurality of hinges. The hinges candidates may be detected by computer software using automated software-based image analysis techniques. Adjusted second image hinge locations may then be calculated based at least in part on the projected second image hinge locations and candidate second image hinge locations. The adjusted second image hinge locations may be used to determine physical locations of the fixator and anatomical structure segments in three-dimensional space, which may be used to determine manipulations to the fixator for deformity correction.

Abstract: The present disclosure relates to systems and methods for image splicing. The systems and methods may acquire a first image and a second image, determine a plurality of first feature points in a first region of the first image, determine a plurality of second feature points in a second region of the second image, then match the plurality of first feature points with the plurality of second feature points to generate a plurality of point pairs. Based on the plurality of point pairs, a third region on the first image and a fourth region on the second image may be determined. Finally, a third image may be generated based on the first image and the second image, wherein the third region of the first image may overlap with the fourth region of the second image in the third image.

Abstract: A deep learning-based digital staining method and system are disclosed that enables the creation of digitally/virtually-stained microscopic images from label or stain-free samples based on autofluorescence images acquired using a fluorescent microscope. The system and method have particular applicability for the creation of digitally/virtually-stained whole slide images (WSIs) of unlabeled/unstained tissue samples that are analyzes by a histopathologist. The methods bypass the standard histochemical staining process, saving time and cost. This method is based on deep learning, and uses, in one embodiment, a convolutional neural network trained using a generative adversarial network model to transform fluorescence images of an unlabeled sample into an image that is equivalent to the brightfield image of the chemically stained-version of the same sample.

Abstract: A method includes identifying a first image that is captured at a first time. The method also includes segmenting the first image into a plurality of first image portions. The method also includes identifying a second image that is captured at a second time. The method also includes segmenting the second image into a plurality of second image portions. The method also includes comparing one of the plurality of first image portions and a corresponding one of the plurality of second image portions. The method also includes determining a difference between the first image and the second image based at least partially upon the comparison. The method also includes transforming the first image into a transformed first image based at least partially upon the difference.

Abstract: Computer-implemented method, systems, computer program products include a processor(s) that obtains input data comprising an image and image data: coordinates (points) defining a region of interest in the image. The processor(s) determines a non-duplicative sequential order for navigating the coordinates. The processor(s) defines a set of points bordering the region of interest: points in the image and points comprising intersection points between two coordinates with an edge of a pixel in the image. The processor(s) determines a combined area of sub pixel subregions in each pixel that includes points in the set of points by: identifying the points of the set, generating one or more sub pixel subregions, and utilizing the sub pixel subregions to determine the combined area of sub pixel subregions.

Abstract: Various embodiments of the present disclosure provide for accelerated segmentation for reverse engineering of integrated circuits. In one example, an embodiment provides for receiving an SEM image for an integrated circuit, performing filtering and binarization with respect to the SEM image, extracting information associated with filter sizes for the filtering, extracting signatures related to a distribution for background pixels and foreground pixels of the SEM image, extracting respective distance to mean signatures for the background pixels and the foreground pixels, and segmenting the SEM image based at least in part on the filter sizes and the respective distance to mean signatures to generate a segmented image for the integrated circuit.

Abstract: A method includes receiving a first image. The method also includes detecting a plurality of edges in the first image. The method also includes connecting the edges. The method also includes identifying a contour in the first image based at least partially upon the connected edges. The method also includes determining a convex hull of the contour. The method also includes generating a second image comprising the convex hull.

Abstract: The techniques described herein relate to methods, apparatus, and computer readable media configured to determining a two-dimensional (2D) profile of a portion of a three-dimensional (3D) point cloud. A 3D region of interest is determined that includes a width along a first axis, a height along a second axis, and a depth along a third axis. The 3D points within the 3D region of interest are represented as a set of 2D points based on coordinate values of the first and second axes. The 2D points are grouped into a plurality of 2D bins arranged along the first axis. For each 2D bin, a representative 2D position is determined based on the associated set of 2D points. Each of the representative 2D positions are connected to neighboring representative 2D positions to generate the 2D profile.

Abstract: A system and method for generating and displaying contours. In some embodiments, the method includes: calculating a first threshold, based on a first target point in a first slice; and generating a first contour, the first contour being a boundary of a first region, the first region being a region within which: the elements of the first slice are greater than the first threshold, or the elements of the first slice are less than the first threshold.

Abstract: The present disclosure describes an image processor of a display device. The image processing includes a pre-processor, a segmentation processor, and a correction processor. The pre-processor performs spatial filtering on an input image signal and output a line image signal. The segmentation processor classifies a class of the line image signal and outputting a segmentation signal representing the class. The correction processor corrects the segmentation signal based on the line image signal. Additionally, the correction processor detects a class boundary of the segmentation signal, detects an edge within an edge region of the line image signal corresponding to the class boundary, and corrects the segmentation signal based on the detected edge.

Abstract: The invention provides an image segmentation method and an electronic device. The image segmentation method includes the following steps. Regression analysis is performed on a first gray-scale image to obtain a residual image having an object backbone area. A pixel value of each pixel in the object backbone area is defined as an average gray-scale value of the object backbone area in the residual image, and a second gray-scale image having the object backbone area is generated. It is recursively determined whether a residual polarity of each adjacent pixel adjacent to edge pixels of the object backbone area in the residual image is the same as a residual polarity of the corresponding edge pixel, and whether a pixel value of each adjacent pixel is greater than a first threshold, so as to expand the object backbone area in the second gray-scale image, which is extracted as a target object.

Abstract: An image processing apparatus includes a memory configured to store first region detection information of a first frame; and a processor. The processor is configured to: identify a first pixel region corresponding to the first region detection information from a second frame, perform region growing processing on the first pixel region based on an adjacent pixel region that is adjacent to the first pixel region, obtain second region detection information of the second frame, based on the region growing processing, and perform image processing on the second frame based on the second region detection information.

Abstract: A focus following method based on a motion gravity center, a storage medium and a photographing system are provided. The focus following method includes the following steps: S1, acquiring video data captured by a camera within an initial time period; S2, acquiring video data with only a living object within the initial time period; S3, acquiring a motion gravity center image from the video data with only the living object according to a motion gravity center library; and S4, calculating an initial motion direction and an initial motion speed of the motion gravity center image within the initial time period. According to the present disclosure, the calculation amount is greatly reduced, the complex data is simplified, focus following based on the motion gravity center under poor hardware conditions can be can achieved, and requirements of low-cost and mass-production hardware can be satisfied.

Abstract: A machine-learning (ML) architecture for determining three or more outputs, such as a two and/or three-dimensional region of interest, semantic segmentation, direction logits, depth data, and/or instance segmentation associated with an object in an image. The ML architecture may output these outputs at a rate of 30 or more frames per second on consumer grade hardware.

Abstract: This disclosure relates generally to system and method for forecasting location of target in monocular first person view. Conventional systems for location forecasting utilizes complex neural networks and hence are computationally intensive and requires high compute power. The disclosed system includes an efficient and light-weight RNN based network model for predicting motion of targets in first person monocular videos. The network model includes an auto-encoder in the encoding phase and a regularizing layer in the end helps us get better accuracy. The disclosed method relies entirely just on detection bounding boxes for prediction as well as training of the network model and is still capable of transferring zero-shot on a different dataset.