44 results about "Visual patterns" patented technology

The invention provides a full-slice image classification method and system based on multi-branch attention and random instance masks and a medium, and the method comprises the steps: obtaining a full-slice digital pathological image, carrying out the image segmentation, and obtaining a plurality of instances; performing feature extraction based on a feature extraction network to obtain an instance feature sequence; carrying out parallel analysis on the instance feature sequence based on a multi-branch attention mechanism, executing a random Top-K instance mask operation, and carrying out weighted summation on the instance feature sequence based on the attention distribution of each attention branch; aggregating the packet level feature representations of all the attention branches, generating a comprehensive feature representation of the full-slice digital pathological image, and obtaining a classification prediction result; by setting a plurality of parallel attention branches, different branches are promoted to actively learn and capture a plurality of different visual modes existing in the full-slice digital pathological image, so that tumor heterogeneity can be effectively represented, and the generalization ability of full-slice digital pathological image classification is improved.

The invention discloses a traffic flow prediction method based on cross-modal interactive geographic image coding, and relates to the technical field of intelligent traffic systems and data processing. The method comprises the following steps: firstly, acquiring historical traffic data and a road network geographic image, and respectively constructing an embedded representation containing space-time periodicity and extracting node-level visual features; secondly, through a hybrid cross attention mechanism, utilizing a learnable global visual token as an abstract agent, compressing visual features and performing cross-modal alignment with the dynamic space-time representation to generate an enhanced space-time representation; meanwhile, a visual relation mode is constructed based on local visual patches between the nodes, and refining is carried out through a general relation matrix; and finally, dynamically fusing time, space and visual relation characteristics by using an adaptive gating mechanism, and predicting future traffic flow. Visual modes such as road geometry can be effectively captured, the problem of dynamic and static heterogeneous mode alignment is solved, and the accuracy of traffic prediction is remarkably improved while the calculation efficiency is guaranteed.

The application discloses a kind of unmanned plane vision enhancement system and method based on binocular rotating polarizer, it is related to computer vision and optical imaging technical field, system includes binocular camera module, rotatable polarizer assembly that can be integrally moved into / out of optical path, mechanism for driving its lifting and rotation and main control unit.The method comprises: after moving polarizer into optical path, in a single acquisition cycle, it is driven to rotate at high speed to multiple preset angles and synchronously acquires binocular image sequence;Process image sequence to calculate Stokes parameters and construct three-dimensional polarization feature space in combination with depth information;The feature space is input into multimodal fusion neural network with original image, and enhanced image, semantic segmentation map and optimized depth map are output.The application realizes the quick and flexible switching of polarization mode and ordinary vision mode, and significantly improves the real-time vision perception and enhancement capability of unmanned plane in underwater, strong reflection and other complex environments through software and hardware depth collaboration.

A system comprising an output device, a brain oscillation monitor, and one or more processors. The output device is configured to output an audio signal for audio stimulation of the patient, and a visual pattern for video stimulation of the patient. The brain oscillation monitor is configured to generate feedback indicative of a response to the audio stimulation and the visual stimulation. The one or more processors are configured to determine a target frequency for the audio stimulation and the visual stimulation. The one or more processors are further configured to cause constructive interference of the audio stimulation and the visual stimulation output by the output device at the target frequency, by modifying at least one of a phase of the visual pattern or an amplitude of onsets for the visual pattern, relative to the audio signal, according to the feedback from the brain oscillation monitor.

The invention discloses a vehicle visual mode switching method, device and equipment and a computer program product. The method comprises the steps that the real-time vehicle distance, the threat fuzzy coefficient, the real-time vehicle speed, the traffic jam condition and the dynamic balance level of a current vehicle and the driving habit style of a driver are obtained; inputting the real-time vehicle distance, the threat fuzzy coefficient, the real-time vehicle speed, the traffic jam condition, the dynamic balance level and the driving habit style into a pre-trained visual mode decision model to obtain a target visual mode; switching the visual mode of the current vehicle to a target visual mode, and controlling a corresponding target visual sensor to operate; wherein the target visual mode is one of a vector-level visual mode, a contour-level visual mode and a high-precision visual mode. On the premise of ensuring the driving safety of the vehicle, the computing power resource consumption of the vehicle is optimized, so that the performance of vehicle hardware is improved, the service life of the vehicle hardware is prolonged, and the method can be applied to the technical field of intelligent driving.

A tamper-evident item authentication label may include a substrate layer that includes a first substrate side and a second substrate side, a first adhesive layer, a second adhesive layer, a first unique indicium, a second unique indicium, and a plurality of cuts across the first unique indicium and the second unique indicium extending through at least a portion of a thickness of the substrate layer. The adhesive layers may be deposited on the substrate sides and arranged to fuse the tamper-evident item authentication label with a first and second surface. The first unique indicium may be deposited on the substrate layer and encoding a first unique identifier that links the tamper-evident item authentication label to one or more records in a remote database system. The second unique indicium may be deposited on the substrate layer. The second unique indicium may include a stochastically created visual pattern.

An eye exam can be performed using an electronic device in a virtual environment to determine vision corrective measures based on vision correction simulation. The electronic device can execute a visual assessment application for displaying a user interface to create a 3D virtual environment corresponding to a field of view of a user associated with the electronic device. The electronic device can render a visual pattern in the field of view and apply a vision correction filter to the visual pattern. The electronic device can obtain a set of user response data captured by a plurality of sensors in response to the visual pattern and determine whether the set of user response data satisfy a response quality criterion. Filter parameters of the vision correction filter can be dynamically adjusted based on the set of user response data until the set of user response data satisfy the response quality criterion.

A patient's visual health can be evaluated via a virtual reality (VR) system, which can include a VR headset in electronic communication with a computing device. The computing device can cause a phenomenon (such as a visual test or an image) to be displayed on the screens of the VR headset. Using varying combinations of sensors, cameras, probes, and microphones, the VR headset collects data about the patient as she perceives and responds to the phenomenon. Optionally, the computing device can alter the phenomenon and analyze the patient's perception and responses to evaluate the patient for macular degeneration. Optionally, the computing device can alter the phenomenon based on the patient's perception and responses.

The utility model provides a small-size folding light path visual training device which comprises a front shell, a middle frame and a rear shell which are connected in sequence, the front shell is provided with a lens device, and a visual training module, a left darkroom, a right darkroom, a flexible FPC and a reflection lens are arranged between the middle frame and the rear shell; the visual training module, the reflecting lens, the left darkroom, the right darkroom and the lens device jointly form a light path system. According to the visual training device with the small-size folding light path, the problems that an existing device is large in size, close in imaging position and poor in experience are solved, the visual training device is heavy to wear, the left eye and the right eye can see the same visual pattern through the spirally-arranged LED light sources, the left eye and the right eye share one visual cylinder, the size is saved, and the imaging distance is increased.

A system for producing an image to design a product can include a processor and a memory. The memory can store a regularizing module, a blending module, a denoising module, and a communications module. The regularizing module can produce a regularized image of a denoised image of an interpolation of a first diffused image and a second diffused image. The regularized image can be regularized with respect to a visual pattern. The blending module can: (1) determine a blending weight and (2) produce, based on the blending weight, a blended image of the denoised image and the regularized image. The denoising module can denoise the blended image to produce the image to design the product. The communications module can cause the image to be sent to a computer-aided design system to design the product.

The invention relates to an unmanned aerial vehicle intelligent path planning method and system based on visual signals, and belongs to the field of path planning. The method comprises the following steps: optimizing a neural network through progressive pruning; fuzzy coordinates and scene types are provided for the unmanned aerial vehicle, the unmanned aerial vehicle flies to a fuzzy coordinate area, and when the unmanned aerial vehicle is close to the fuzzy coordinate area, the unmanned aerial vehicle is switched to a pure vision mode to perform environment perception through a neural network, and a three-dimensional semantic map is constructed; performing real-time path planning based on the three-dimensional semantic map to obtain a flight path; the method comprises the following steps: performing autonomous flight along a flight path, when approaching an operation target, performing autonomous operation to obtain real-time monitoring data, storing the real-time monitoring data in airborne storage equipment, and returning the real-time monitoring data after returning. And an unmanned aerial vehicle intelligent path planning system with high robustness, high autonomy and high task adaptability is constructed.

The invention relates to the technical field of training equipment, and discloses a multi-modal stimulation mouse rich environment training device, which comprises a box body, a training cage and a visual stimulation assembly, the training cage and the visual stimulation assembly are arranged in the box body; the visual stimulation assembly comprises a face gear and a plurality of visual moving assemblies, the face gear is provided with a meshing end face facing the training cage, and the outer periphery of the training cage is meshed with the meshing end face; the multiple visual moving assemblies are arranged on the meshing end face at intervals. The visual movement assembly is movably arranged in the radial direction of the meshing end face, and the visual stimulation assembly further comprises a resistance adjusting mechanism used for adjusting resistance of the visual movement assembly during radial movement. The visual moving assembly is movably arranged in the radial direction of the face gear, and visual patterns seen by a mouse not only translate, but also continuously jump, shake or pulse, so that the cognitive load of training is greatly increased.

The application belongs to the technical field of computer vision, and provides a visual Transformer self-supervised learning method and system based on multi-dimensional relationship modeling; the method uses self-relationship modeling on spatial dimensions and channel dimensions, uses different image transformation to process images to obtain different views of the images; the different views of the images are processed by a teacher network and a student network respectively to obtain feature maps; the feature map extracted by the student network is further processed by a convolution layer; a self-relationship matrix of the feature map in the spatial dimensions and the channel dimensions is calculated by point multiplication; the difference between the relationship matrices extracted by the teacher network and the student network is calculated as a loss function, and the derivative of the loss function with respect to the network parameters is used to update the network parameters; compared with the existing self-supervised learning method which only considers the features of visual patterns, the method simultaneously considers the correlation of visual patterns in the spatial and channel dimensions, and can significantly improve the accuracy of image recognition, semantic segmentation, target detection, instance detection and the like.

The invention discloses an unmanned aerial vehicle vision enhancement system and method based on a binocular rotating polarizer, and relates to the technical field of computer vision and optical imaging, and the system comprises a binocular camera module, a rotatable polarizer assembly capable of integrally moving in / out of a light path, a mechanism for driving the rotatable polarizer assembly to lift and rotate, and a main control unit. The method comprises the following steps: after moving a polarizer into a light path, driving the polarizer to rotate at a high speed to a plurality of preset angles in a single acquisition period and synchronously acquiring a binocular image sequence; processing the image sequence to calculate a Stokes parameter and constructing a three-dimensional polarization feature space in combination with depth information; and inputting the feature space and the original image into a multi-modal fusion neural network, and outputting an enhanced image, a semantic segmentation image and an optimized depth image. According to the invention, rapid and flexible switching between a polarization mode and a common visual mode is realized, and the real-time visual perception and enhancement capability of the unmanned aerial vehicle in complex environments such as underwater and strong reflection is significantly improved through deep cooperation of software and hardware.

The invention discloses a man-machine cooperation continuous image compression method and system based on a hybrid expert adapter, and the technical scheme is that the method comprises the steps: generating and preprocessing a general feature data set and a downstream task data set, and building a basic coding and decoding network; constructing a hybrid expert adapter containing a public expert module and a task expert module in parallel, wherein each expert internally comprises a bottleneck type lightweight structure and spatial frequency domain double branches; a man-machine cooperation decoupling architecture is constructed, two-stage training is adopted, and a parameter isolation strategy is implemented in the second stage; the modes can be dynamically switched during reasoning, wherein the mixed expert adapter needs to be removed in the human eye vision mode, and the corresponding experts are dynamically activated according to task identification in the machine vision mode. According to the method, the human eye vision quality and the machine vision task performance are considered, multi-task zero-forgetting continuous learning is supported, and the downstream machine vision task analysis precision is improved.

An automatic pick-and-place system is provided, including a feeding device, a visual pattern formed on a fixed part of the feeding device, a manipulator, and a processing unit. An electronic component can be held between the fixed part and a gripping part of the feeding device. A visual module on the manipulator captures an image of the visual pattern, and the processing unit calculates the coordinate value of the electronic component. Thus, the manipulator can move to the target position and pick up the electronic element according to the coordinate value.

The invention discloses an input method, a related device, electronic equipment and a storage medium, and the input method comprises the steps: obtaining a plurality of mapping rules in response to a first character sequence of a selected word on a screen in an input process; wherein the mapping rule comprises a second character sequence and a visual pattern associated with the second character sequence; selecting a target rule from the plurality of mapping rules based on a matching result between the first character sequence and the second character sequence in each mapping rule; and carrying out replacement display on the current cursor based on the visual pattern in the target rule. According to the scheme, the feedback and individuation requirements in the input process can be met at the same time.

System (104) for controlling a vehicle (102), wherein the system (104) comprises the following: an electronic processor (200) designed to: To take, via a camera (112), a first image; Determine, within the first image, of road traffic factors which road lines (402A-402E) and at least one road user near the vehicle (102) are included; Generating, based on sensor information from one or more sensors (114) of the vehicle (102), a second image (400) depicting an environment surrounding the vehicle, wherein the second image (400) includes the road traffic factors; where the road lines (402A-402E) in the second image (400) have a visual pattern or color which indicates the designated direction of travel; Determine, based on the detected road traffic factors and the second image (400), a predicted trajectory (408A-408D) of the road user near the vehicle; wherein the trajectory (408A-408D) in the second image (400) is represented by a gradient of a visual pattern or color, where the lightest portion is the earliest position and the darkest portion is the latest position of the road user near the vehicle (102); Generating a steering command for the vehicle (102) based on the

The invention provides an image-enhanced molecular representation learning method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence and drug discovery. The image enhancement molecular representation learning method mainly comprises the following steps: learning topological features from a molecular structure by adopting a graph neural network encoder; performing self-supervised pre-training on the graph encoder through a contrast learning strategy to obtain general molecular characterization; utilizing a pre-training image encoder to extract visual mode features of the molecular two-dimensional image; designing a fusion module based on an attention mechanism, dynamically integrating information of two modes of graph structure and image vision, and generating an enhanced molecular representation vector; and completing classification or regression prediction of molecular attributes through the enhanced representation. By implementing the image enhancement molecular representation learning method and equipment provided by the invention, the molecular structure and visual information can be fused, and the accuracy and generalization ability of molecular property prediction are improved.

The utility model relates to a kind of rotating video conversion cosmetic bottles, including bottle and bottle cap, pattern is equipped on the bottle body surface of bottle, bottom is fixed with base at the bottom of bottle, bearing is installed in base, cylinder is sleeved on the outside of bottle, grating is equipped on cylinder;Bottom is equipped with bottom disc at the bottom of bottle, bottom disc is connected with bearing sleeve and forms bottom disc rotatable structure;Bottom disc is equipped with clamping groove, the lower side of cylinder is inserted into the clamping groove of bottom disc, the bayonet of cylinder lower side and the clamping block in clamping groove are engaged butt joint, constitute the structure of cylinder rotation with bottom disc, when bottle body is rotated relative to cylinder, the pattern on bottle body surface is converted into interactive dynamic visual pattern by grating on cylinder.Utilize grating principle, by hollow stripe cylinder shell or transparent film cylinder that printing has grating stripe in bottle body sleeve, realize that bottle body presents quite different visual picture, color or information under different rotation angle, when user rotates cosmetic bottle, user sees dynamic pattern on bottle body.

An eye exam can be performed using an electronic device in a virtual environment to provide a corrective vision prescription covering a field of view. The electronic device can execute a visual assessment application, e.g., by displaying a user interface to create a 3D virtual environment. The electronic device can partition a field of view displayed on the user interface into a plurality of regions. For each of the plurality of regions in the field of view, successively, the electronic device can render a respective visual pattern in the respective region, obtain a user response to the respective visual pattern, and adjust a respective vision correction filter to the respective visual pattern based on the user response. Respective vision correction filters corresponding to the plurality of regions are combined to determine a prescription of an eyewear for a user associated with the electronic device.

The invention discloses a food image recognition and calorie recognition method based on a convolutional neural network, and belongs to the field of image recognition. A food image lacks a fixed semantic mode, and local detail features of the food image present a multi-scale characteristic. Limited by fixed damage size and receptive field, DCL cannot extract multi-scale discriminative features. The method comprises the following steps: preprocessing image data in a selected data set; training a convolutional neural network by using the fused multi-stage convolutional feature pyramid; an ECA attention module is added, and different feature maps are reweighted to enhance a discriminative visual mode; capturing local cross-channel interaction; extracting features by using a convolutional neural network which is trained and is added with an attention mechanism; carrying out feature fusion; a target detection module and a heat prediction module are added; and performing target detection and calorie calculation on the input to-be-detected food image. The multi-scale feature learning ability of the network is enhanced by using the feature pyramid, and the food identification precision is improved.

A system may be configured for implementing targeted attacks on deep reinforcement learning-based autonomous driving with learned visual patterns. In some examples, processing circuitry receives first input specifying an initial state for a driving environment and user configurable input specifying a target state. Processing circuitry may generate a representative dataset of the driving environment by performing multiple rollouts of the vehicle through the driving environment, including performing an action for the vehicle from the initial state with variable strength noise added to determine a next state for each rollout resulting from the action. Processing circuitry may train an artificial intelligence model to output a next predicted state based on the representative dataset as training input. In such an example, processing circuitry outputs from the artificial intelligence model, an attack plan against the autonomous driving agent to achieve the target state from the initial state.

The invention discloses a bearing surface defect detection method and system and a related device, and the method comprises the steps: rubbing the surface of a negative mold, and generating a rubbing pattern: attaching a flexible rubbing layer to the surface of the negative mold, applying uniform pressure to the back surface of the flexible rubbing layer, and after the flexible rubbing layer stays for a preset time, separating the flexible rubbing layer; and obtaining a rubbing pattern based on the flexible rubbing layer. The method comprises the following steps of: adopting a negative mold reverse structure matched with the shape lines on the surface of a bearing to be detected, re-engraving the microstructure characteristics of the surface of the bearing, converting the three-dimensional surface appearance into a two-dimensional visual pattern through physical rubbing, reserving the spatial distribution characteristics of surface defects, and establishing a preset standard rubbing pattern database, so that the defects are conveniently compared and positioned, and the accuracy of the surface defect detection is improved. And the requirement of bearing surface defect detection is effectively met.

The invention relates to an image classification method and device based on a side suppression attention mechanism and electronic equipment, and the method comprises the steps: constructing an image classification model based on the side suppression attention mechanism, and carrying out the training through employing a neuromorphic data set, and obtaining a trained image classification model based on the side suppression attention mechanism; inputting a to-be-processed neuromorphic image into the trained image classification model based on the lateral inhibition attention mechanism to obtain an image classification result; according to the method, secondary features and background information are actively inhibited, so that the network is more focused on a key visual mode, and the feature identification capability and robustness of the model are improved under the condition that the parameter quantity is not excessively increased; according to the method, the effectiveness and generalization ability of a side suppression attention mechanism in an image processing task provide a new thought and method support for a low-power-consumption and high-efficiency bionic vision calculation model.

A digital image radial pattern decoding system is described. In one example, an unfolded digital image is formed by the radial pattern decoding system by unfolding a radial pattern in a digital image. An inflated digital image is then generated by the radial pattern decoding system by upsampling the unfolded radial pattern. A grid pattern is determined by the radial pattern decoding system based on the inflated digital image. A radial pattern cell is then generated based on a reverse transform of the grid pattern. A visual pattern is generated by the radial pattern decoding system based on the radial pattern cell.

The embodiment of the invention relates to the technical field of vehicle-mounted information entertainment and intelligent cabin man-machine interaction, and discloses a vehicle-mounted cabin music immersive visual linkage method, device and equipment and a medium, and the method comprises the steps: obtaining an album cover image and an audio data stream of currently played music in response to a music playing instruction; performing dominant hue extraction on the cover image to obtain at least one theme color; and displaying a dynamic visual graph synchronized with the music rhythm on a vehicle-mounted main screen based on the theme color and the audio data stream. By applying the technical scheme of the invention, the music album cover can be used as a unified visual emotion source, and the deviation between the subjectively set visual change and the music rhythm is avoided through real-time sound and picture data fusion and synchronous rendering, so that the visual response can be accurately matched with the core characteristics such as the strength and the rhythm of the music, and the music quality is improved. And the dynamic response accuracy of the system is improved.

The application provides a diversity inductive bias learning method for a generalizable large model, comprising constructing an input pair of images according to a complete prompt description of a training sample set; using a generalizable large model to extract image features and text features of the input pair; and using cosine similarity loss to supervise training for the same input pair. In order to bridge the semantic gap between language and visual patterns, the application proposes a text-level inductive bias, provides detailed information for each category by supplementing many LLM-generated descriptions in the prompt text; in order to enable the model to well capture the inductive bias, a phrase adapter is designed for the text encoder to explicitly explore the connection between adjacent words; a space adapter is designed for the image encoder to allow the model to see more local relationships and details; the application reduces overfitting by optimizing the level inductive bias; this is achieved through a dynamic training strategy, which enables the model to learn different degrees of fitting state.

A computer-implemented method and system for validating vehicle damage detection utilizes symmetry-based analysis of opposing vehicle sides. The method comprises receiving images from opposing sides of a vehicle, detecting a damage region in one image, and computing deep learning feature vectors for the damage region and a corresponding region in the opposite image. These feature vectors represent learned visual patterns that discriminate between damage and normal vehicle features. A similarity measure is computed between the feature vectors, and the detected damage is validated based on this measure. The system includes sensors for image capture, processors, and memory storing instructions to execute the method. This approach leverages vehicle symmetry to reduce false positives, compensate for environmental variations, and improve damage detection accuracy. The method can adapt to asymmetric vehicle positioning and varying environmental conditions, providing robust performance in real-world scenarios.