Abstract: A method for managing image data in an automotive lighting device. This method includes the steps of providing an image pattern, dividing the image pattern in rows or columns of pixels, and providing for each row pattern, a plurality of linear segments, each linear segment is characterized by two values. The data of the linear segments is compressed, and the compressed data is sent to the light module. The invention also provides an automotive lighting device for performing the steps of such a method.

Abstract: A method and system including: defining a geographic area; receiving a plurality of images; determining a plurality of image points; partitioning the geographic area into a plurality of image regions based on the plurality of image points; and stitching the plurality of images into a combined image based on the plurality of image regions.

Abstract: An image processing method and a device configured to implement the same are disclosed. The method comprises: obtaining optical input from a hybrid imaging device, wherein an obtained optical input comprises a first component and a second component that temporally corresponds to the first component; wherein the first component of the obtained optical input corresponds to a first temporal resolution, while the second component of the obtained optical input corresponds to a second temporal resolution higher than that of the first component; performing image restoration operation on a first subset of the first component of the obtained optical input in accordance with data from the second component of the obtained optical input; and performing image fusion operation to generate fused image data from an output of the image restoration operation and a second subset of the first component of the obtained optical input.

Abstract: A method for acquiring motion vectors includes: obtaining a number of related motion vectors of a current block, wherein the number of related motion vectors are motion vectors of a related block of the current block, a prediction mode of the related block is the same as a current prediction mode of the current block, an encoding has been completed; determining weights of at least part of the number of related motion vectors; and calculating a weighted motion vector of the current block based on at least part of the number of related motion vectors and the weights of at least part of the plurality of related motion vectors.

Abstract: System, device, and method for improved encoding and enhanced compression of images, videos, and media content. A method includes: (a) receiving a source image, and analyzing its content on a pixel-by-pixel basis, and classifying each pixel as either (I) a pixel associated with Photographic content, or (II) a pixel associated with Non-Photographic content; (c) generating a pixel-clusters map that indicates (i) clusters of pixels that were classified as Photographic content, and (ii) clusters of pixels that were classified as Non-Photographic content; (d) generating a composed image, by: (d1) applying a first encoding technique, particularly lossy encoding, to encode pixel-clusters that were classified as Photographic content; (d2) applying a second, different, encoding technique, particularly lossless encoding, to encode pixel-clusters that were classified as Non-Photographic content.

Abstract: An image processing apparatus performs external recognition for driving assistance or automated driving of a vehicle based on a captured image obtained from a photographing apparatus capturing an image requiring distortion correction. The image processing apparatus: moves and sets a partial region to be a target of distortion correction processing according to a rule, and applies the distortion correction processing to the set partial region of the captured image to acquire a partial image corrected in distortion; detects a road mirror around the vehicle; and changes the rule so that a partial region corresponding to a predetermined direction of the vehicle is preferentially set as a target of the distortion correction processing when a road mirror is detected.

Abstract: Systems, apparatuses, and methods are described for 3D luminescence imaging, by identifying a preferred optical pair and optimizing a scanned image using the preferred optical pair. An optimal filter pair may be selected from a list of two or more optical filters. An acceptable threshold of information may be obtained using a subset of the list of two or more optical filters (e.g., an optimal filter pair). An imaging device may be configured with the optimal filter pair to produce a pair of luminescence images of a target sample. In addition, luminescence images may be pre-processed to reduce the time-cost of conventional processing techniques of luminescence images. One or more computing devices may generate initial prior data based on a pair of luminescence images. An output may include one or more output luminescent sources that have been refined and/or optimized from the initial prior data.

Abstract: A system or process may generate a summarization of multimedia content by determining one or more salient moments therefrom. Multimedia content may be received and a plurality of frames and audio, visual, and metadata elements associated therewith are extracted from the multimedia content. A plurality of importance sub-scores may be generated for each frame of the multimedia content, each of the plurality of sub-scores being associated with a particular analytical modality. For each frame, the plurality of importance sub-scores associated therewith may be aggregated into an importance score. The frames may be ranked by importance and a plurality of top-ranked frames are identified and determined to satisfy an importance threshold. The plurality of top-ranked frames are sequentially arranged and merged into a plurality of moment candidates that are ranked for importance. A subset of top-ranked moment candidates are merged into a final summarization of the multimedia content.

Abstract: A given segment of computer code is obtained and modified to produce one or more inefficient versions of the given segment of computer code in comparison to the given segment of computer code. A code parse tree is generated for the given segment of computer code and each inefficient version of the given segment of computer code. Model embeddings are generated based on the generated code parse trees and a diffusion model is trained based on the generated model embeddings.

Abstract: An image processing method and apparatus, an electronic device, and a storage medium, relating to the technical field of image processing. The method includes: acquiring a first image and a second image, wherein a resolution of the second image is greater than a resolution of the first image; determining difference information between a target pixel in the second image and a reference pixel corresponding to the target pixel point in the first image; and acquire a target image with the same resolution as the second image by applying an image differencing process to a predetermined image with the same resolution as the first image based on the difference information. This method can obtain the target image based on the difference value information, and improve image quality.

Abstract: A multi-scale image processing algorithm for enhancing contrast in an electronic representation of an image represented by a) decomposing said digital image into a set of detail images at multiple resolution levels and a residual image at a resolution level lower than said multiple resolution levels, b) processing at least one pixel of said detail images, c) computing a processed image by applying a reconstruction algorithm to the residual image and the processed detail images, said reconstruction algorithm being such that if it were applied to the residual image and the detail images without processing, then said digital image or a close approximation thereof would be obtained.

Abstract: A method for defending against an adversarial sample in image classification includes: denoising, by an adversarial denoising network, an input image to acquire a reconstructed image; acquiring, by a target classification model, a predicted category probability distribution of the reconstructed image; acquiring, by the target classification model, a predicted category probability distribution of the original input image; calculating an adversarial score of the input image, and determining the input image as an adversarial sample or a benign sample according to a threshold; outputting a category prediction result of the reconstructed image if the input image is determined as the adversarial sample; and outputting a category prediction result of the original input image if the input image is determined as the benign sample. A system for defending against an adversarial sample in image classification, and a data processing terminal are further provided.

Abstract: An image processing method includes: acquiring a first image containing a target object; inputting the first image into an image processing model to obtain a second image, the second image being a mask image of the target object in the first image, a value for each pixel in the second image being in a range of 0 to 1, inclusive ([0, 1]), and the range of 0 to 1, inclusive ([0, 1]) indicating a degree of relation between each pixel in the second image and a pixel in the target object; fusing the first image and a background image according to the second image to obtain a fused image; and providing a first interface and displaying the fused image on the first interface.

Abstract: Layouts for elements of a flowchart or flow diagram may use recursive algorithms that trace along each parent-child pathway to ensure that child elements always occur after parent elements in a horizontal spacing. Vertical spacing requirements may then be calculated for each element, along with a vertical space above a center line and a vertical space below a center line for each element. These vertical spacing requirements may then be used by a recursive algorithm to assign vertical positions to each of the elements. The vertical spacing may ensure that lanes created for each branch element are not crossed by later elements in the process flow, which ensures that connections between elements may be clearly distinguished from each other.

Abstract: A computer-implemented method is provided. The computer-implemented method includes inputting a low-resolution image and a plurality of high-resolution images into a feature extractor, the low-resolution image and the plurality of high-resolution images including images of a target object; obtaining, by the feature extractor, feature maps of the low-resolution image and the plurality of high-resolution images; comparing similarities between the feature maps of the low-resolution image and the plurality of high-resolution images; obtaining selected feature maps of one or more selected high-resolution images of the plurality of high-resolution images most similar to the low-resolution image; inputting the selected feature maps into a generator to output a repair image; enhancing the low-resolution image using a pre-processing image enhancing process to generate an enhanced image; and morphing the repair image with the enhanced image.

Abstract: Provided is a learning device that acquires computational imaging information of a computational imaging camera that captures an image with blurring; acquires a normal image captured by a normal camera that captures an image without blurring or an image with blurring smaller than that of the computational imaging camera, and a correct answer label assigned to the normal image; generates an image with blurring based on the computational imaging information and the normal image; and performs machine learning using the image with blurring and the correct answer label to create an image identification model for identifying an image captured by the computational imaging camera.

Abstract: An arithmetic part of an arithmetic processing device includes: a filter processing part that has a multiplier and a first adder and performs filter processing; a second adder that performs cumulative addition processing that cumulatively adds all of the results of the filter processing as executed in N parallel; a non-linear conversion part that performs non-linear arithmetic processing on the result of the cumulative addition processing; a pooling processing part that performs pooling processing on the result of the non-linear arithmetic processing; and an arithmetic control part that controls the filter processing part, the second adder, the non-linear conversion part, and the pooling processing part.

Abstract: An image processing method includes displaying a first key frame image obtained from a sequence frame video stream; obtaining a first foreground image and a first background image obtained after foreground and background separation processing is performed on the first key frame image; obtaining a second foreground image obtained after foreground image processing is performed on a first target object in the first foreground image; obtaining a third background image obtained after background repair processing is performed on the first background image using a second background image, where the second background image is a background image included in a second key frame image obtained by photographing a target scene before the first key frame image is obtained; and obtaining a first key frame image obtained after foreground and background compositing processing is performed on the second foreground image and the third background image.

Abstract: A system and method are disclosed for low-light image enhancement using denoising preprocessing with wavelet decomposition AI-based techniques to enhance image quality of low-light images. Subsampled images are created from a raw input image. A wavelet decomposition process is performed on each subimage to create multiple frequency domain subimages. Each frequency domain subimage is input into a corresponding neural network. The output of each corresponding network is input to an inverse wavelet module. The output of the inverse wavelet module is a denoised image that is input to an image signal processing pipeline, where additional processing may be performed on the denoised image.

Abstract: This application discloses an image processing method and apparatus, and relates to the field of image processing technologies, to help optimize a color, a contrast, or a dynamic range of an image, so that an optimized image can be more objective, and robustness is improved. The method is applied to a terminal including a first camera and a second camera. The method includes: when an ISO of the first camera in a current photographing environment is greater than a first threshold, capturing a first image for a first scenario in the current photographing environment; capturing a second image for the first scenario; and optimizing the first image based on the second image to obtain a third image. An image style of the second image is better than that of the first image.

Abstract: Image captured through a non-rectilinear lens may exhibit distortions. The distortions may be reduced by warping the image. However, warping the image may degrade the fidelity of the image. The warped image may be enhanced to increase the fidelity of the image. The enhancement may be applied to the portions of the image that were degraded from the warping.

Abstract: Machine vision techniques for determining a region of interest (ROI) are disclosed herein. An example implementation includes a computing device for executing an application, the application operable to: (1) receive a first image; (2) set an first ROI of the first image to a field of view (FOV) of the first image; (3) determine a barcode within the first ROI; (4) determine a bounding box of the barcode; (5) form a second ROI based on the bounding box; (6) receive a second image; and (7) set an ROI of the second image to be the second ROI of the first image.

Abstract: There are a medical image processing apparatus and a medical image processing method that are capable of maintaining the estimation accuracy of noise intensity even though a system noise ratio rises. A medical image processing apparatus includes a difference image generating unit that divides projection data obtained by applying radiation to an examinee to create a difference image between tomographic images reconstructed for every divided piece of projection data; a local variance computing unit that computes local variance in the difference image; and a noise estimation unit that corrects the local variance using a correction function found in advance to estimate noise intensity of a tomographic image of the examinee. The correction function includes system noise.

Abstract: Methods, systems, and apparatuses for adaptive processing of video content to remove noise, such as film grain noise, without substantially affecting visual presentation quality are described herein. A computing device may determine a plurality of film grain parameters associated with film grain noise present within one or more portions of a content item. The computing device may determine at least one encoding parameter based on the plurality of film grain parameters. The computing device may encode the content item based on the at least one encoding parameter. The computing device may send an encoding message to at least one user device/client device, which may in turn use the encoding message to decode the content item.

Abstract: One or more computing devices, systems, and/or methods for defect detection are provided. An image, depicting an object for evaluation to determine whether the object has a defect, is inputted into a segmentation model to identify an object region of interest of the object. An object region area of the object region of interest is calculated. A convex hull area of a convex hull encompassing the object region of interest is calculated. A ratio of the object region area to the convex hull area is determined. The ratio is compared to a threshold to determine whether the object has the defect or does not have the defect.

Abstract: A rotating wind turbine blade optical inspection devices and systems. The system for inspection, having: a camera; a rotatable device designed to controllably rotate camera along axis, not camera's lens axis of rotational symmetry; a laser distance sensor or a dynamic vision sensor and a computer system; the camera, the rotatable device and the laser distance sensor or the dynamic vision sensor operably connected with the computer system; wherein the camera is adapted to be mounted on the rotatable device; the distance or the vision sensors' optical axis is directed at a set distance from the camera's optical axis; wherein the computer system: (i) to receives a signal from the distance or vision sensors; (ii) to determines camera's activation time based on the signal from the distance sensor or vision sensors and a pre-set adjustable delay time; (iii) to activate the camera acquiring an image of a blade area inspected.

Abstract: An apparatus includes a processing node of a distributed computing platform. The processing node communicates with other processing nodes over one or more networks. The processing node may receive frames of point clouds at a processing node of a distributed computing platform, determine a subset of the frames as key frames based at least in part on distances travelled between captures of the respective frames, and allocate tasks of processing the key frames to processing subnodes based at least in part on estimated processing demands of the key frames and processing capabilities of each of the processing subnodes.

Abstract: A structural analysis computing device may generate a proposed insurance claim and/or generate a proposed insurance quote for an object pictured in a three-dimensional (3D) image. The structural analysis computing device may be coupled to a drone configured to capture exterior images of the object. The structural analysis computing device may include a memory, a user interface, an object sensor configured to capture the 3D image, and a processor in communication with the memory and the object sensor. The processor may access the 3D image including the object, and analyze the 3D images to identify features of the object—such as by inputting the 3D image into a trained machine learning or pattern recognition program. The processor may generate a proposed claim form for a damaged object and/or a proposed quote for an uninsured object, and display the form to a user for their review and/or approval.

Abstract: Disclosed are a positioning method, an electronic device, and a storage medium. The method includes: converting, according to an intrinsic parameter matrix of an image acquisition module and depth values of multiple pixel points in a depth image captured by the image acquisition module, pixel coordinates of the multiple pixel points into camera coordinates; and matching the camera coordinates of the respective pixel points with target world coordinates of multiple voxels in a pre-established point cloud map, and obtaining a positioning result of the image acquisition module; wherein the target world coordinates of the respective voxels are obtained by updating initial world coordinates of the respective voxels according to multiple sample image pairs, and a sample image pair comprises a two-dimensional sample image and a depth sample image.

Abstract: An image processing method. The method includes: An electronic device obtains N images, where the N images have a same quantity of pixels and a same pixel location arrangement, and N is an integer greater than 1; the electronic device obtains, based on feature values of pixels located at a same location in the N images, a reference value of the corresponding location; the electronic device determines a target pixel of each location based on a reference value of the location; and the electronic device generates a target image based on the target pixel of each location.

Abstract: A display device including a display; one or more processors; and a memory storing one or more programs configured to be executed by the one or more processors, the one or more programs including instructions for receiving an input image to be displayed on the display; selecting a look-up table (LUT) curve stored in the memory for the received input image; performing a histogram on the received input image; applying an offset gain to the selected LUT curve based on the histogram to generate an offset LUT curve; performing tone mapping of the input image through the offset LUT curve; and displaying the tone mapped image on the display.

Abstract: Aspects and implementations provide for mechanisms of detection and decoding of barcodes in images. The disclosed techniques include estimating dimensions of a module of a barcode based on geometric characteristics of a barcode image, forming hypotheses that group modules into barcode symbols, and assessing viability of formed hypotheses. Various operations of the techniques may involve the use of neural networks, including estimation of module dimensions and assessment of groupings of modules into lines and lines into barcode symbols. The techniques may be used for decoding of barcodes captured in images of unfavorable conditions, including blur, perspective, sub-optimal lighting, barcode deformation, and the like. The techniques may be applied to decoding linear one-dimensional barcodes, two-dimensional barcodes, and stacked linear barcodes.

Abstract: An image processing apparatus includes a composition unit configured to composite a plurality of captured images to generate a composite image, a recording control unit configured to record the captured images or the composite image as a recorded image in a recording unit, and a determination unit configured to determine an imaging mode or a method of composition performed by the composition unit, in which the recording control unit records type data determined on the basis of a result of the determination in association with the recorded image that is captured or composited according to the imaging mode or the method of composition determined by the determination unit.

Abstract: A printer-equipped camera includes an instant film pack loading unit in which an instant film pack, a first displaying unit of which a displaying surface is arranged to face an exposure surface of an instant film in the instant film pack through the exposure opening with respect to the instant film pack loaded in the instant film pack loading unit, and that exposes the instant film by displaying an image, a light exit direction restriction member that is included on the displaying surface of the first displaying unit and restricts a light exit direction of light from each pixel of the first displaying unit to a constant range, an imaging unit that captures a subject image, a second displaying unit that displays an image to an outside, a print instruction unit that provides a print instruction for the image displayed on the second displaying unit, and a displaying control unit that exposes the exposure surface of the instant film by displaying the image on the first displaying unit.

Abstract: In some embodiments, a method receives a text prompt and executes a text encoder on the text prompt to generate an embedding representation. A set of base images is generated based on the embedding representation and parameters of a base image generation model. A high resolution model is executed to upsample one or more base images in the set of base images based on parameters of the high resolution model to generate a set of final images. The method ranks the set of base images or the set of final images using reward values that are generated by a reward model. The reward model is trained using human input that provided feedback on a quality of generated images using the base image generation model and the high resolution model. One or more final images are output based on the ranking in response to the text prompt.

Abstract: A method, including receiving a computerized tomography (CT) image of voxels of a subject's head, and analyzing the image to identify respective locations of the subject's eyes in the image, so defining a first line segment joining the respective locations. The method includes identifying a voxel subset overlaying bony sections of the head, lying on a second line segment parallel to the first line segment and on a third line segment orthogonal to the first line segment. A magnetic tracking system configured to measure positions on the subject's head is activated, and a probe, operative in the system, is positioned in proximity to the bony sections to measure positions of a surface of the head overlaying the bony sections. A correspondence between the positions and the voxel subset is formed, and a registration between the CT image and the magnetic tracking system is generated in response to the correspondence.

Abstract: A system and method for improving a video characteristic of a video stream is described. According to various implementations of the invention, a changed region between a later-in-time image frame and an earlier-in-time image frame and an unchanged region between such two image frames are determined. A new improvement to the video characteristic is determined and applied to the changed region of the later-in-time image frame. A prior improvement to the video characteristic that was determined for the earlier-in-time image frame is applied to the unchanged region of the later-in-time image frame.

Abstract: An electronic apparatus is provided. The electronic apparatus includes: a memory configured to store an image; a sensor part including at least one sensor; a projection part including a projection lens configured to output the image to a projection surface; and a processor configured to acquire distance information from the electronic apparatus to the projection surface through the sensor part, acquire output size information of the image based on the acquired distance information, acquire projection surface information corresponding to a bend of the projection surface through the sensor part, acquire movement information of the projection lens based on the output size information of the image and the projection surface information, and control the projection part to output the image based on the movement information.

Abstract: An image processing method includes: determining a brightness value of an image; and enhancing a brightness of the image when the brightness value of the image is less than an image brightness threshold. Enhancing the brightness of the image includes: determining each pixel of the image as a target pixel; determining a brightness enhancement value of the target pixel based on the brightness value of the image, an initial brightness value of the target pixel, and initial brightness values of neighboring pixels of the target pixel; and using the brightness enhancement value as an enhanced brightness value of the target pixel.

Abstract: An image processing apparatus and method is provided and includes one or more processors and one or more memory devices that store instructions. When the instructions are executed by the one or more processors, the one or more processors are configured to perform operations including obtaining a first image which is obtained by capturing based on a first shooting parameter, generating, by image processing on the first image, a second image corresponding to a second shooting parameter which is different from the first shooting parameter and using at least the generated second image as a training image for a model that is used in noise estimation processing for an input image.

Abstract: An AI-assisted automatic labeling system and a method thereof are disclosed. The method comprises steps: selecting images from microscopic images as candidate images, using a pre-labeling module to automatically label cells in the candidate images, and dividing the labeled images into training data and verification data; using a training module and the training data to train a basic model; using a verification module to verify and modify the basic model, wherein the verification module respectively verifies at least one cell area and at least one background area of the verification data to converge the basic model and form an automatic labeling model; using the automatic labeling model to automatically label cells in redundant images of the microscopic images. The basic model trained by the present invention can use few labeled images to perform regressive training and verification and then automatically labels the redundant images accurately and efficiently.

Abstract: A method includes obtaining a reference image and a target image each representing an environment containing moving features and static features. The method also includes determining an object mask configured to mask out the moving features and preserves the static features in the target image. The method additionally includes determining, based on motion parallax between the reference image and the target image, a static depth image representing depth values of the static features in the target image. The method further includes generating, by way of a machine learning model, a dynamic depth image representing depth values of both the static features and the moving features in the target image. The model is trained to generate the dynamic depth image by determining depth values of at least the moving features based on the target image, the object mask, and the static depth image.

Abstract: An image processing method, including the following steps: receiving an original image signal, wherein the original image signal includes a plurality of original pixel values; identifying an edge of at least one object in the original image signal to generate a contour image signal; and correcting the original image signal according to the contour image signal to generate an enhanced image signal.

Abstract: Disclosed herein are methods and apparatus for making a determination about a cataract in an eye in ambient lighting conditions.

Abstract: Automated inventory management and material (or container) handling removes the requirement to operate fully automatically or all-manual using conventional task dedicated vertical storage and retrieval (S&R) machines. Inventory requests Automated vehicles plan their own movements to execute missions over a container yard, warehouse aisles or roadways, sharing this space with manually driven trucks. Automated units drive to planned speed limits, manage their loads (stability control), stop, go, and merge at intersections according human driving rules, use on-board sensors to identify static and dynamic obstacles, and human traffic, and either avoid them or stop until potential collision risk is removed. They identify, localize, and either pick-up loads (pallets, container, etc.) or drop them at the correctly demined locations. Systems without full automation can also implement partially automated operations (for instance load pick-up and drop), and can assure inherently safe manually operated vehicles (i.

Abstract: Evaluating dose performance of a radiographic imaging system with respect to image quality using a phantom, a channelized hotelling observer module as a model observer, and a printer, a plaque, or an electronic display includes scanning and producing images for a plurality of sections of the phantom using the radiographic imaging system, wherein the plurality of sections represent a range of patient sizes and doses and wherein the sections of the phantom contain objects of measurable detectability. Also included is analyzing the images to determine detectability results for one or more of the contained objects within the images of the plurality of sections of the phantom, wherein the analyzing includes using a channelized hotelling observer (CHO) module as a model observer; and displaying, via the printer, the plaque, or the electronic display, a continuous detectability performance measurement function using the determined detectability results.

Abstract: Motion contaminated magnetic resonance (MR) images for training an artificial neural network to remove motion artifacts from the MR images are difficult to obtain. Described herein are systems, methods, and instrumentalities for injecting motion artifacts into clean MR images and using the artificially contaminated images for machine learning and neural network training. The motion contaminated MR images may be created based on clean source MR images that are associated with multiple physiological cycles of a scanned object, and by deriving MR data segments for the multiple physiological cycles based on the source MR images. The MR data segments thus derived may be combined to obtain a simulated MR data set, from which one or more target MR images may be generated to exhibit a motion artifact. The motion artifact may be created by manipulating the source MR images and/or controlling the manner in which the MR data set or the target MR images are generated.

Abstract: Apparatus, systems and methods for adaptively reducing blocking artifacts in block-coded video are disclosed. In one implementation, a system includes processing logic at least capable of deblock filtering at least a portion of a line of video data based, at least in part, on edge information and texture information to generate at least a portion of a line of deblocked video data, and an image data output device responsive to the processing logic.

Abstract: A training data generation device performs the processes of: acquiring first correct answer data that is correct answer data with respect to each training image of a training image group, the training image group including plural training images to be given the correct answer data, the training image group being divided into plural image subgroups, the first correct answer data being generated in a different situation for each of the plural image subgroups; generating a discriminator with respect to the each of the plural image subgroups, the discriminator being learned based on the each of the plural image subgroups and the first correct answer data corresponding to the each of the plural image subgroups; acquiring a discrimination result outputted by the discriminator for each of the plural training images by inputting the each of the plural training images to the discriminator which is generated based on at least one image subgroup other than an image subgroup to which the each of the training images belon

Abstract: In various examples, systems and methods are disclosed that preserve rich spatial information from an input resolution of a machine learning model to regress on lines in an input image. The machine learning model may be trained to predict, in deployment, distances for each pixel of the input image at an input resolution to a line pixel determined to correspond to a line in the input image. The machine learning model may further be trained to predict angles and label classes of the line. An embedding algorithm may be used to train the machine learning model to predict clusters of line pixels that each correspond to a respective line in the input image. In deployment, the predictions of the machine learning model may be used as an aid for understanding the surrounding environment—e.g., for updating a world model—in a variety of autonomous machine applications.