Abstract: In some examples, a controller can include a processing resource and a memory resource storing instructions to cause the processing resource perform object detection on a frame of a plurality of frames of a video signal, generate a bounding box around an object in the frame in response to the object detection detecting the object in the frame, perform a backlight adjustment of the frame, and generate an output frame having adjusted pixel values for each pixel from the backlight adjustment.

Abstract: A method of detecting a change of ratio of occlusion of a tracked object in a video sequence. For each of a plurality of image frames, a bounding box of the tracked object is determined and for each pair of successive image frames, an intersection over union (IoU) of a first bounding box in a first image frame and a second bounding box in a second image frame is calculated. Similarly, for a further pair of successive image frames, a further IoU of a first bounding box in a first image frame of the further pair of successive image frames and a second bounding box in a second image frame of the further pair of successive image frames is calculated. If the further IoU differs from the calculated IoUs by more than a threshold amount, the ratio of occlusion of the tracked object has changed.

Abstract: A method of forming an optical neural network for processing an input object image or optical signal that is invariant to object transformations includes training a software-based neural network model to perform one or more specific optical functions for a multi-layer optical network having physical features located in each of the layers of the optical neural network. The training includes feeding different input object images or optical signals that have random transformations or shifts and computing at least one optical output of optical transmission and/or reflection through the optical neural network using an optical wave propagation model and iteratively adjusting transmission/reflection coefficients for each layer until optimized transmission/reflection coefficients are obtained.

Abstract: According to an embodiment of the disclosure, an image processing device, which is capable of detecting a distortable pattern that is likely to be distorted, includes a block luminance acquirer configured to divide a plurality of pixels constituting an input image into a plurality of blocks and acquire luminance for each of the plurality of blocks, a pattern detector configured to analyze a luminance correlation between the plurality of blocks and detect a distortable pattern, and a luminance corrector configured to correct the luminance of the plurality of pixels or bypass the plurality of pixels according to whether the distortable pattern is detected.

Abstract: In various embodiments a training application trains convolutional neural networks (CNNs) to reduce reconstruction errors. The training application executes a first CNN on a source image having a first resolution to generate a downscaled image having a second resolution. The training application executes a second CNN on the downscaled image to generate a reconstructed image having the first resolution. The training application computes a reconstruction error based on the reconstructed image and the source image. The training application updates a first learnable parameter value included in the first CNN based on the reconstruction error to generate at least a partially trained downscaling CNN. The training application updates a second learnable parameter included in the second CNN based on the reconstruction error to generate at least a partially trained upscaling CNN.

Abstract: Neural network operation apparatus and method are provided. The neural network operation apparatus includes: one or more processors; and memory storing instructions configured to cause the one or more processors to: generate an upsampled tensor by copying pixels, of a unit of data, based on a scale factor for upsampling; and generate, based on the scale factor, a neural network operation result by performing a pooling operation on the upsampled tensor.

Abstract: A crankshaft shape inspection method includes: acquiring three-dimensional point cloud data of a surface of a crankshaft S; superposing the three-dimensional point cloud data on a surface shape model of the crankshaft S; moving the three-dimensional point cloud data superposed on the surface shape model to match with a coordinate system used when the crankshaft S is machined; generating an estimated machined surface, which is the surface after machining of a predetermined machining portion of the crankshaft S, in the coordinate system used when the crankshaft S is machined; and calculating a distance between machining portion point cloud data extracted from the three-dimensional point cloud data moved and the estimated machined surface generated and determining a machining stock of the crankshaft S to be insufficient based on the calculated distance.

Abstract: Systems and methods for performing a medical analysis task using a trained machine learning based task network are provided. Input medical data is received. A medical analysis task is performed using a trained machine learning based task network based on the input medical data. Results of the medical analysis task are output. The trained machine learning based task network is trained by: receiving unannotated training medical data; generating weakly-supervised labels for the unannotated training medical data using one or more trained machine learning based supervised learning networks; training the machine learning based task network for performing the medical analysis task based on 1) the unannotated training medical data, 2) self-supervised labels for the unannotated training medical data learned via self-supervised learning, and 3) the generated weakly-supervised labels for the unannotated training medical data; and outputting the trained machine learning based task network.

Abstract: A method of predicting a semiconductor yield includes receiving wafer level data generated by measuring a plurality of wafers, generating a plurality of virtual chips corresponding to the plurality of wafers based on the wafer level data, mapping a test result of the plurality of wafers to the plurality of virtual chips, computing a defect rate of the plurality of virtual chips according to defects based on a result of the mapping, and computing a defect index of the equipment based on the defect rate.

Abstract: Methods and systems for generating a feature map from an image is disclosed. The vision system includes a vision model or processing the image to generate the feature map according a neural network. The vision model includes a first convolutional block for downsampling a set of image data to obtain a first stage convoluted data; a second convolutional block for downsampling the first stage convoluted data to obtain a second stage convoluted data, wherein one or both of the first convolutional block and the second convolutional block is a mobile convolution block (MBConv) that includes: a first Gaussian Error Linear Unit (GELU) layer, a depth-wise convolution (DWConv) layer having, and a resizing convolution layer; and a transformer block (TFB) generating the feature map from the second stage convoluted data.

Abstract: Systems and methods are provided for enabling detection of prohibited objects using three-dimensional imaging data. The system obtains a three-dimensional representation of at least part of an item. The 3D representation includes multiple points with associated intensity values determined based on data acquired by an imaging device. A cluster of points corresponding to an object of interest is identified within the 3D representation. From this cluster, data indicating the position and the orientation of the object is determined. The data indicating the position and the orientation of the object is used to generate a two-dimensional representation of the object with an orientation which is chosen to facilitate detectability of the object as a prohibited item by a detection algorithm (such as a machine learning model).

Abstract: A method for matching a three-dimensional first image of at least one discrete entity with a three-dimensional second image of the at least one discrete entity is provided. The at least one discrete entity includes a biological sample and a plurality of constituent parts of a marker. The method includes: generating a first representation of the marker from the first image; generating a second representation of the marker from the second image; and based upon the representations matching, matching the first image with the second image; or based upon the representations not matching, rejecting the match. Generating the representations includes determining vectors from at least one reference item to at least some of the constituent parts of the marker, determining for the vectors at least one value of a property, and generating the representations of the marker based on a frequency of the at least one value of the property.

Abstract: A rendering system includes a rendering unit and a correction unit. The rendering unit renders, based on a result of detection by a detection system, a marker corresponding to a location of a target. The detection system is installed in a moving vehicle for the purpose of detecting the target. The correction unit corrects, based on correction data, the location of the target in accordance with the result of detection by the detection system and thereby determines a location of the marker to be rendered by the rendering unit. The correction data is obtained based on at least traveling information about a traveling condition of the moving vehicle.

Abstract: Systems and methods disclosed herein relate to using a machine-learning model to process an input of a subject's eye and to predict a current or future visual acuity of the subject. The subject may have been diagnosed with age-related macular degeneration. The predicted current or future visual acuity may be used to (for example) facilitate diagnosing the subject (e.g., with a particular type of age-related macular degeneration), facilitate identifying a treatment strategy for the subject, and/or facilitate designing a clinical study.

Abstract: A method for inspecting an object includes determining guide image data of the object from a determined orientation, the guide image data including a guide image pixel array and a pixel property for at least one guide image pixel in the guide image pixel array. The method also includes receiving inspection image data indicative of an inspection image and associating the inspection image data with the guide image data with a processor of a computing device. Additionally, the method includes determining a property of the object based on the guide image data and the associated inspection image data.

Abstract: A method includes receiving, from an image sensor, an image, identifying, by a first neural network, a plurality of locations-of-interest within the image, and generating, by the first neural network, a first classification label for each location-of-interest of the plurality of locations-of-interest. The method also includes extracting, from the image, a plurality of image chips derived from the plurality of locations-of-interest and generating, by a second neural network, a second classification label for each image chip of the plurality of image chips. The method further includes determining an identification of a set of targets within the image using the plurality of locations-of-interest, the first classification label for each location-of-interest of the plurality of locations-of-interest, the plurality of image chips, and the second classification label for each image chip of the plurality of image chips, and transmitting the identification of the set of targets within the image.

Abstract: A method and device for performing a perception task are disclosed. The method and device incorporate a dense regression model. The dense regression model advantageously incorporates a distortion-free convolution technique that is designed to accommodate and appropriately handle the varying levels of distortion in omnidirectional images across different regions. In addition to distortion-free convolution, the dense regression model further utilizes a transformer that incorporates an spherical self-attention that use distortion-free image embedding to compute an appearance attention and uses spherical distance to compute a positional attention.

Abstract: A method of image processing includes: determining a first feature, wherein the first feature has a dimensionality D1; determining a second feature, wherein the second feature has a dimensionality D2 and is based on an output of a feature extraction network; generating a third feature by processing the first feature, the third feature having a dimensionality D3; generating a guidance by processing the second feature, the guidance having the dimensionality D3; generating a filter output by applying a deep guided filter (DGF) to the third feature using the guidance; generating a map based on the filter output; and outputting a processed image based on the map.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer-readable media that generate joint-based segmentation masks for digital objects portrayed in digital videos. In particular, in one or more embodiments, the disclosed systems utilize a video masking model having a pose tracking neural network and a segmentation neural network to generate the joint-based segmentation masks. To illustrate, in some embodiments, the disclosed systems utilize the pose tracking neural network to identify a set of joints of the digital object across the frames of the digital video. The disclosed systems further utilize the segmentation neural network to generate joint-based segmentation masks for the video frames that portray the object using the identified joints. In some cases, the segmentation neural network includes a multi-layer perceptron mixer layer for mixing visual features propagated via convolutional layers.

Abstract: This disclosure provides systems, methods, and devices for processing and aligning sensor data features for navigation. In a first aspect, a method is provided that includes determining, based on received sensor data, a first set of features for an area surrounding the vehicle. A second set of features for the area surrounding the vehicle may be determined based on an occupancy map for the area surrounding the vehicle. A third set of features may be determined that align the first set of features with the second set of features. The third set of features may align each of at least a subset of the second set of features with at least one corresponding feature from the first set of features. Other aspects and features are also claimed and described.