Abstract: A method and a device for self-supervised learning, a storage medium, and an electronic device are provided. The method includes: organizing real points in one column along a vertical direction into a pillar; determining a predicted point in a next frame; determining a first loss term based on a minimum distance among distances between predicted points in the next frame and real points in the next frame, and generating a loss function including the first loss term; and performing self-supervised learning processing based on the loss function. A pillar motion parameter representing motion of a real point is determined with the pillar as a unit, so as to enhance correlation between point clouds. Self-supervised learning can be realized in a case of no precise correspondence between the predicted point and the real point, and training is performed based on a large number of unlabeled point clouds.

Abstract: Methods and systems for automated assessment of spermatogenesis. Embodiments disclosed herein relate to drug development and testicular toxicity in safety evaluation studies, and more particularly to automatic assessment of spermatogenesis through a staging of seminiferous tubules using Artificial Intelligence/deep learning methods. A method disclosed herein includes detecting the seminiferous tubules by analyzing a testes tissue specimen and mapping the seminiferous tubules to detect and segment germ cells. The method includes classifying the seminiferous tubules into respective stages based on the segmented germ cells. The method further includes categorizing the seminiferous tubules into a normal category and an abnormal category based on the segmented germ cells. The method further includes categorizing the testes tissue specimen into the normal category and the abnormal category based on the classification of the seminiferous tubules for toxicity analysis.

Abstract: According to the techniques of this disclosure, a method includes capturing, using a camera system of a vehicle, at least one image of an occupant of the vehicle, determining, based on the at least one image of the occupant, a location of one or more eyes of the occupant within the vehicle, and determining, based on the at least one image of the occupant, an eye gaze vector. The method may also include determining, based on the eye gaze vector, the location of the one or more eyes of the occupant, and a vehicle data file of the vehicle, a region of interest from a plurality of regions of interests of the vehicle at which the occupant is looking, wherein the vehicle data file specifies respective locations of each of the plurality of regions of interest, and selectively performing, based on the region of interest, an action.

Abstract: Aspects of the subject matter disclosed herein include methods, systems, and other techniques for training, in a first phase, an object classifier neural network with a first set of training data, the first set of training data including a first plurality of training examples, each training example in the first set of training data being labeled with a coarse-object classification; and training, in a second phase after completion of the first phase, the object classifier neural network with a second set of training data, the second set of training data including a second plurality of training examples, each training example in the second set of training data being labeled with a fine-object classification.

Abstract: A system for image segmentation is provided. The system may obtain a target image including an ROI, and segment a preliminary region representative of the ROI from the target image using a first ROI segmentation model corresponding to a first image resolution. The system may segment a target region representative of the ROI from the preliminary region using a second ROI segmentation model corresponding to a second image resolution. At least one model of the first and second ROI segmentation models may at least include a first convolutional layer and a second convolutional layer downstream to the first convolutional layer. A count of input channels of the first convolutional layer may be greater than a count of output channels of the first convolutional layer, and a count of input channels of the second convolutional layer may be smaller than a count of output channels of the second convolutional layer.

Abstract: A biometric capture device is operative to adjust one or more environmental parameters to enhance a range (e.g., distance) at which a biometric may be captured from a subject. For example, a sample biometric capture device may be, include, or otherwise incorporate a retinal or iris scanner configured to capture an image of the retina or iris (e.g., a biometric) when the retina or iris is illuminated by infrared light. Generally, the amount of infrared light required to accurately image the retina or iris increases with the distance of the subject's retina or iris from the image capture device. The biometric capture device may capture a facial image using a first image sensor, identify a face in the facial image, capture an iris image using a second image sensor guided by the facial image, and identify a person using the iris image.

Abstract: Methods and systems for training computer-aided condition detection systems. One method includes receiving a plurality of images for a plurality of patients, some of the images including an annotation associated with a condition; iteratively applying a first deep learning network to each of the images to produce a segmentation map, a feature map, and an image-level probability of the condition for each of the images; iteratively applying a second deep learning network to each feature map produced by the first network to produce a plurality of outputs; training the first network based on the segmentation map produced for each image; and training the second network based on the output produced for each of the patients. The second network includes a plurality of convolution layers and a plurality of convolutional long short-term memory (LSTM) layers. Each of the outputs includes a patient-level probability of the condition for one of the patients.

Abstract: Systems and methods monitor a workspace for safety purposes using sensors distributed about the workspace. The sensors are registered with respect to each other, and this registration is monitored over time. Occluded space as well as occupied space is identified, and this mapping is frequently updated. Based on the mapping, a constrained motion plan of machinery can be generated to ensure safety.

Abstract: System and method for diagnosing a disease using image classification is provided. The method includes receiving diagnostic images associated with a subject. The diagnostic images comprise a first image and a second image captured at a first instance of time and at a second instance of time, respectively. The method further includes extracting feature maps from the first image and the second image, using a first pipeline and a second pipeline of stacked convolutional layers of a temporal convolutional neural network (CNN) model respectively. The method further includes identifying relevant feature maps from the feature maps using temporal CNN layer of the temporal CNN model, generating a concatenated feature map based on the relevant feature maps, determining an image class corresponding to the second image using fully connected layers of the temporal CNN model, and diagnosing the disease associated with the subject based on the image class.

Abstract: Embodiments described herein provide various examples of a machine-learning-based visual-haptic system for constructing visual-haptic models for various interactions between surgical tools and tissues. In one aspect, a process for constructing a visual-haptic model is disclosed. This process can begin by receiving a set of training videos. The process then processes each training video in the set of training videos to extract one or more video segments that depict a target tool-tissue interaction from the training video, wherein the target tool-tissue interaction involves exerting a force by one or more surgical tools on a tissue. Next, for each video segment in the set of video segments, the process annotates each video image in the video segment with a set of force levels predefined for the target tool-tissue interaction. The process subsequently trains a machine-learning model using the annotated video images to obtain a trained machine-learning model for the target tool-tissue interaction.

Abstract: A method of generating a font database, and a method of training a neural network model are provided, which relate to a field of artificial intelligence, in particular to a computer vision and deep learning technology. The method of generating the font database includes: determining, by using a trained similarity comparison model, a basic font database most similar to handwriting font data of a target user in a plurality of basic font databases as a candidate font database; and adjusting, by using a trained basic font database model for generating the candidate font database, the handwriting font data of the target user, so as to obtain a target font database for the target user.

Abstract: The disclosure provides a system that may receive an identification of a first patient; may retrieve, based at least on the identification of the first patient, first eye identification information that includes a first plurality of iris structures associated with a first eye of the first patient; may determine a second plurality of iris structures of an eye of a current patient; may determine if the second plurality of iris structures match the first plurality of iris structures; if the second plurality of iris structures match the first plurality of iris structures, may provide an indication that the first eye has been correctly identified; and if the second plurality of iris structures do match the first plurality of iris structures, may provide an indication that the first eye has not been correctly identified.

Abstract: An apparatus includes a memory and processor. The memory stores document categories, text generated from an image a physical document page, and a machine learning algorithm. The text includes errors associated with noise in the image. The machine learning algorithm is configured to extract features associated with natural language processing and features associated with the errors from the text. The machine learning algorithm is also configured to generate a feature vector that includes the first and second pluralities of features, and to generate, based on the feature vector, a set of probabilities, each of which is associated with a document category and indicates a probability that the physical document from which the text was generated belongs to that document category. The processor applies the machine learning algorithm to the text, to generate the set of probabilities, identifies a largest probability, and assigns the image to the associated document category.

Abstract: A method (and imaging system) for planning a medical intervention on patient is provided that involves an intracavity probe and an imaging dataset of the patient. A view-type is selected from a defined set of view-types. A virtual field of view of the intracavity probe corresponding to the selected view-type is determined. A virtual intracavity image is rendered for display, wherein the virtual intracavity image is based upon the imaging dataset of the patient and the virtual field of view of the intracavity probe. The virtual field of view can be based upon segmentation of an intracavity path of the probe and at least one anatomical structure or possibly based on user input. In embodiments, the virtual field of view can be based upon probe parameters computed in accordance with a pre-defined set of rules for the selected view-type. The probe parameters can be computed by evaluation of a cost function expressed by the pre-defined set of rules for the selected view-type.

Abstract: A system is provided for a distributed server network with embedded image decoder as a chain code program runtime event. In particular, the system may comprise a distributed computing network comprising one or more decentralized nodes, each of which may store a separate copy of a distributed data register. The system may further comprise one or more specialized nodes which receive, assess, and analyze user input data, where the one or more specialized nodes may include a client identity node comprising an embedded image decoder which may be configured to analyze image portions of the user input data. Once the image data has been analyzed, the client identity node may convert the image data into a text format for storage within the distributed register.

Abstract: A system comprises one or more processors. The system further comprises memory having computer readable instructions stored thereon that, when executed by the one or more processors, cause the system to: obtain a reference tree of nodes and linkages for a reference three dimensional volumetric representation of a branched anatomical formation in a reference state; based on the reference tree, generate a reference three dimensional geometric model; based on measurements from a sensor system, detect deformation of the branched anatomical formation; based on the detected deformation, obtain a deformed tree of nodes and linkages; and generate a three dimensional deformation field that represents the detected deformation of the branched anatomical formation.

Abstract: An exemplary system, method, and computer-accessible medium for determining a position or a characteristic of a target(s) for a transcranial magnetic stimulation (TMS) treatment of a patient(s) can be provided, which can include, for example, receiving imaging information of a portion(s) of a head of the patient(s), and determining the position or the characteristic of the target(s) for the TMS treatment of the patient(s) based on the imaging information. The imaging information can be magnetic resonance imaging information. The imaging information can include information regarding a brain and a skull of the patient(s). The position or the characteristic of the target(s) can be determined by identifying (i) the skull, and (ii) a parcel in a section(s) of a brain of the patient(s). The parcel can a dorsolateral prefrontal cortex (DLPFC) parcel. The DLPFC parcel can be identified using a parcellation procedure, which can be a human connectome pipeline procedure.

Abstract: Systems, devices, media and methods are presented for a human pose tracking framework. The human pose tracking framework may identify a message with video frames, generate, using a composite convolutional neural network, joint data representing joint locations of a human depicted in the video frames, the generating of the joint data by the composite convolutional neural network done by a deep convolutional neural network operating on one portion of the video frames, a shallow convolutional neural network operating on a another portion of the video frames, and tracking the joint locations using a one-shot learner neural network that is trained to track the joint locations based on a concatenation of feature maps and a convolutional pose machine. The human pose tracking framework may store, the joint locations, and cause presentation of a rendition of the joint locations on a user interface of a client device.

Abstract: A biometric capture device is operative to adjust one or more environmental parameters to enhance a range (e.g., distance) at which a biometric may be captured from a subject. For example, a sample biometric capture device may be, include, or otherwise incorporate a retinal or iris scanner configured to capture an image of the retina or iris (e.g., a biometric) when the retina or iris is illuminated by infrared light. Generally, the amount of infrared light required to accurately image the retina or iris increases with the distance of the subject's retina or iris from the image capture device. The biometric capture device may capture a facial image using a first image sensor, identify a face in the facial image, capture an iris image using a second image sensor guided by the facial image, and identify a person using the iris image.

Abstract: Method and apparatus embodiments can generate a volume fluorescence image of a tooth. Method and apparatus embodiments can project structured light patterns onto a tooth and generate a contour (volume) image of the tooth surface from acquired corresponding structured light projection images; then acquire one or more fluorescence images of the tooth generated under blue-UV illumination. A composite image that shows fluorescence image content mapped to the generated contour image can be transmitted, stored, modified and/or displayed.