86 results about "Scale down" patented technology

Various embodiments of the present disclosure are directed towards an image sensor comprising a pixel with a dual-PD layout for enhanced scaling down. The pixel spans a first integrated circuit (IC) die and a second IC die stacked with the first IC die. The pixel comprises a plurality of photodetectors in the first IC die, and further comprises a plurality of pixel transistors split amongst the first IC die and the second IC die. The plurality of photodetectors are grouped into one or more pairs, each having the dual-PD layout. A DTI structure completely and individually surrounds the plurality of photodetectors, and further extends completely through a substrate within which the plurality of photodetectors are arranged. As such, the DTI structure completely separates the plurality of photodetectors from each other.

The invention relates to the technical field of remote sensing image processing, and particularly discloses a remote sensing image target detection method based on a multi-scale cross-stage network model, and the method comprises the following steps: obtaining a to-be-detected remote sensing image; the method comprises the following steps of: inputting a backbone network after passing through an input layer, extracting multi-scale and multi-level semantic features step by step, and obtaining feature maps of different sizes through multi-scale down-sampling; inputting the feature maps of different sizes into a feature integration network, and performing cross-stage fusion and lightweight optimization to obtain a plurality of fused feature maps; and inputting the plurality of fusion feature maps into a prediction head, and carrying out classification detection to obtain detection results of multiple scales. According to the method, the brand new multi-scale cross-stage network model is constructed, multi-scale feature extraction, fusion and target recognition capabilities are enhanced, accurate detection of multi-scale targets, especially small targets and complex-form targets, in the remote sensing image can be realized, detection precision and reasoning efficiency are both considered, and the method is suitable for remote sensing application scenes such as ocean monitoring and urban planning.

The invention discloses a bridge disease detection method based on a diffusion model and a bitter fish optimization algorithm, and relates to the technical field of bridge detection. The method comprises the following steps: fixing a visual angle and a distance at an easy-to-peel or crack position of a bridge, and collecting and aligning visible light and near-infrared images; carrying out multi-scale downsampling on the image, and carrying out wavelet denoising, brightness correction and texture smoothing; inputting the preprocessing result into an improved diffusion model, weighting the edge during forward diffusion, and reversely generating and applying texture and contour smoothing; the multi-source feature channel and the reconstructed image are combined and input into a deep segmentation network, shadow and stain are eliminated by using a local difference function, and global search is performed on a segmentation threshold, a noise coefficient and the like based on a disease detection rate, a false detection rate and the like by using a bitter fish algorithm; training and correcting the high-noise area again according to the optimal parameters; and uniformly marking disease areas. According to the method, the recognition recall rate of tiny spalling and irregular cracks in an extreme environment can be greatly improved, and the intelligent level and the practical effect of bridge disease detection are improved.

The invention discloses a time sequence self-supervised learning method and system for a rail transit engineering video image, and the method comprises the steps: collecting an urban rail transit engineering construction video, and carrying out the segmentation of the video, and generating segmentation masks, so as to construct a self-supervised pre-training data set; a pre-trained model is selected and initialized; carrying out dynamic frame rate sampling processing on the input video; carrying out random shielding processing on each video frame to generate a corresponding binary mask image; carrying out multi-scale down-sampling on the processed video, and inputting the processed video into a coding-decoding network to complete spatial-temporal feature coding and generation of a reconstructed video and a prediction mask; a loss function is formed by calculating reconstruction loss of a reconstructed video and a real video, dynamic frequency loss based on Fourier domain low-frequency and high-frequency feature consistency and segmentation loss of a prediction mask and a real mask. The problems of insufficient time sequence modeling adaptability and insufficient multi-scale target adaptability are solved.

An embodiment for dynamic optimization of power consumption in storage systems is provided. The embodiment may include receiving historical time-series data from one or more components in a data center. The embodiment may also include predicting one or more workload metrics of the one or more components during a pre-defined time range. The embodiment may further include identifying one or more required resources for the one or more components to handle the predicted one or more workload metrics during the pre-defined time range. The embodiment may also include in response to determining at least one component of the one or more components does not require a resource allocation increase during the pre-defined time range, executing a first action to scale down at least one first resource during the pre-defined time range.

A device may be configured to receive a bitstream. The device may decode the bitstream according to a video coding technique. The device may determine a normalization technique from one or more normalization techniques. The device may perform the normalization technique on the decoded bitstream to generate feature data for input into a neural network. Normalizations techniques may include a combination of a linear normalization techniques, scaling down normalization techniques, and logarithmic normalization techniques.

The invention provides a flow measurement device and method for an upper plate of a reactor core, and the device comprises a measurement model which is provided with a first to-be-measured drain hole which is scaled down according to the first drain hole; the at least one detection mechanism comprises a detection end, and the detection end is suitable for extending into a first to-be-detected drain hole needing to be measured and moving along a preset route, so that the detection mechanism generates a plurality of dynamic pressure data of a plurality of preset point positions, located on the preset route, in the first to-be-detected drain hole; and the flow calculation module is suitable for calculating the measured flow data of the first drain hole to be measured according to the plurality of dynamic pressure data, and performing proportional conversion according to the measured flow data to obtain the actual flow data of the first drain hole corresponding to the first drain hole to be measured. According to the flow measurement device, accurate measurement flow data can be obtained through calculation according to the dynamic pressure data, and then accurate actual flow data of the first drain hole are obtained.

The invention discloses a deep learning point cloud registration method for low-overlapping-rate point clouds. The method comprises the following steps: (1) carrying out multi-scale down-sampling on an input point cloud, calculating a weighted geometric centroid and a normal vector based on an adaptive neighborhood, and constructing a stable local reference point; (2) proposing a centroid geometric structure code, and embedding an attention mechanism by taking an inter-centroid Euclidean distance, a normal included angle and a neighborhood radius ratio as geometric priori to realize feature matching of structure constraint; (3) in a training stage, optimizing overlapping region positioning and point pair matching consistency by adopting a joint loss function formed by circular ring feature loss, overlapping perception loss and matchability loss; and (4) rigid body transformation is estimated by combining a coarse-to-fine matching strategy with a random sampling consistency algorithm, and global high-precision alignment is realized. According to the method, the registration precision and robustness under the conditions of low overlapping rate, noise and shielding are effectively improved, and the method is superior to an existing method in the aspects of a standard data set and real measurement data.

An autoscaling method and apparatus in a Kubernetes cluster, which method and apparatus are applied to the technical field of computers. The method comprises: in historical monitoring data, using an interpolation method to impute missing data between adjacent data points to obtain preprocessed sequence data; performing a fast discrete Fourier transform on the preprocessed sequence data to obtain a spectrogram of original data; analyzing the spectrogram, and identifying candidate periods in the spectrogram; using a preset correlation coefficient calculation function to verify the candidate periods to obtain a dominant period; on the basis of the dominant period and operation data in a cluster, performing peak prediction; and when a predicted peak exceeds a preset autoscaling threshold, triggering a scale-up or scale-down operation corresponding to a system. By means of the present disclosure, the requirement for resources in a period of time in the future can be predicted, and scaling up or scaling down can be performed, so that a more active and accurate resource allocation policy is realized, and the stability of a service is ensured.

The invention discloses an image rain removal method and device, which are used for solving the technical problem of poor image rain removal effect caused by difficulty in accurately balancing rain line removal and background detail recovery in the existing rain removal technology. The method comprises the following steps: acquiring a rain image, performing multi-scale down-sampling on the rain image, and outputting a multi-scale rain image; inputting the image with rain and the multi-scale image with rain into a preset multi-scale dual-domain collaborative rain removal network, wherein the preset multi-scale dual-domain collaborative rain removal network comprises an encoder, a multi-scale space interaction feature fusion module and a decoder; encoding the rain image and the multi-scale rain image by using an encoder, and outputting a plurality of target rain encoding features; carrying out feature fusion on the multiple target rain coding features through a multi-scale space interaction feature fusion module, and outputting target fusion features; and inputting the image with rain, the multi-scale image with rain, the target fusion feature and the target coding feature with rain into a decoder for decoding to generate a target rain-removed image.

The invention relates to the technical field of image processing, and provides a target detection method based on multi-scale feature weighted fusion. The method comprises the following steps: inputting a remote sensing image, and extracting multi-scale features to obtain a multi-level feature map; calculating a first adaptive fusion weight, performing up-sampling alignment on the feature maps of adjacent levels, performing weighted fusion on the feature maps by using the weight, and outputting a first fused feature map; calculating a second adaptive fusion weight, performing multi-scale down-sampling on the first fusion feature map to obtain a down-sampling fusion feature map, performing weighted fusion on the down-sampling fusion feature map by using the second weight, and outputting a second fusion feature; and constructing a target detection network by using a tag allocation strategy of a normalized Wasserstein distance, receiving the second fusion feature by the target detection network to perform target detection, and outputting a detection result, so as to solve the technical problems of insufficient detection precision and non-uniform anchor frame allocation, and achieve the technical effect of improving small target detection precision and tag allocation robustness.

A container-based application management method and apparatus are disclosed. A container-based serverless computing system is configured such that an application instance is allowed to be in one of an online state and a low-power state at runtime. In response to scaling down an application, at least one first application instance of the application in the online state is caused to enter the low-power state; and in response to scaling up the application, at least one second application instance of the application in the low-power state is caused to enter the online state. In this way, not only the cost of the application instance can be reduced, but also rapid elastic scaling can be performed on the application instance.

This invention discloses a method and system for simulating the consistency of droplet generator spreading characteristics between ground and space, used to obtain the spreading characteristics of droplet generators in a vacuum environment. It can equivalently convert ground test results into space test results, solving the problem of equivalent measurement for consistency between ground and space. By scaling down the operating parameters of the jet spreader under rated conditions, it overcomes the adverse effects caused by ground test environment limitations leading to excessively thick droplets that do not match the actual spreading effect in the space environment, thus achieving equivalent measurement for consistency between ground and space. Furthermore, by placing several droplet collectors with the same surface area and material at different locations and measuring the mass increment of multiple droplet collectors, it achieves measurement of liquid film thickness uniformity.

A device may be configured to receive a bitstream. The device may decode the bitstream according to a video coding technique. The device may determine a normalization technique from one or more normalization techniques. The device may perform the normalization technique on the decoded bitstream to generate feature data for input into a neural network. Normalizations techniques may include a combination of a linear normalization techniques, scaling down normalization techniques, and logarithmic normalization techniques.

The utility model discloses a three-phase voltage, frequency and phase sequence detection circuit, and belongs to the technical field of electric power detection. The three-phase voltage, frequency and phase sequence detection circuit comprises an MCU control unit, an A-phase detection circuit, a B-phase detection circuit and a C-phase detection circuit, wherein the A-phase detection circuit, the B-phase detection circuit and the C-phase detection circuit have the same structure. Each detection circuit comprises a voltage sampling circuit and a frequency detection circuit. According to the utility model, the operational amplifier is used for scaling down the input voltage, and the input voltage is skillfully raised to a range which can be accurately sampled by the MCU, compared with a traditional transformer detection method, the size and the weight are greatly reduced, the waveform distortion is extremely small, the characteristics of an original voltage signal are accurately reflected, and the detection precision is greatly improved. The effects of small size, light weight, low cost and high detection precision can be simultaneously achieved.

The present embodiments relate to a write head with an asymmetrical side gap (SG) design. In the asymmetrical design, the main pole can be disposed offset to a central axis of the write head such that the main pole is closer to a first side shield (SS) portion than a second SS portion. The asymmetrical design can be achieved using an ion-beam etching (IBE) or ion-beam deposition (IBD) process. The asymmetrical design can provide a narrower side gap width while mitigating any writability limitations or constraints caused by scaling down the SG.

1. Name of the product in this design: Experimental Demonstration Platform (VMO Model Scaled Down). 2. Purpose of this design: A platform for aircraft experimental demonstrations. 3. The key design feature of this product is its shape. 4. The image or photograph that best illustrates the design's key points: a 3D model.

This invention provides a method for determining the size of a pulse wind tunnel test model, comprising: determining the flow field of the nozzle and test section using numerical simulation to obtain a uniform flow field region S1 in the wind tunnel; then calibrating the uniform flow field region S2 of the nozzle and test section using a cross-shaped frame, comparing the sizes of the two uniform flow fields S1 and S2, and selecting the smaller uniform flow field region S0 as the uniform flow field region S0; scaling down the selected model under the condition of maximum test angle, with the maximum projected area M1 of the model in the direction perpendicular to the airflow occupying 60% to 65% of the cross-sectional area of ​​the uniform flow field S0; integrating the scaled-down model with the nozzle and test section into a single model, numerically simulating the entire flow process under the condition of maximum angle of attack and / or sideslip angle test angle, redetermining the model scale based on the model installation position, the shock wave position at the leading edge of the model, and the coverage of the uniform flow field, and determining the upper limit of the model length through the effective test time and the velocity of the flow field in the test section.

A crystallizer modeling method according to an embodiment of the present disclosure may comprise the steps of: calculating a first stress distribution of a first crystallizer; setting a parameter required for modeling a second crystallizer scaled down from the first crystallizer; calculating a second stress distribution of the second crystallizer on the basis of the set parameter; comparing the second stress distribution with the first stress distribution; determining a parameter for calculating a second stress distribution matching the first stress distribution; and designing the second crystallizer on the basis of the parameter.

This invention discloses a weight inference mapping method for optimizing memristor-based neural networks. First, the weight mapping parameter C is calculated based on the memristor conductance range and the weight range. t And the signal input transformation parameter θ; the weight matrix is ​​transformed according to the weight mapping parameter C t The signal is amplified proportionally and mapped onto a memristor array via an external circuit board. The voltage signal input to the memristor array is first scaled down proportionally according to the signal input conversion parameter θ, and then input into the mapped memristor array for matrix multiplication and addition. A constant resistor array is used to convert the current value output from the memristor array into a voltage signal. This voltage signal is added to the voltage signal at the next time step to obtain the input voltage signal for the next time step, thus initiating the weight inference mapping process for the next cycle. This method solves the problem of timing information loss during the weight inference mapping process, enabling RNN-type algorithms (such as GRU, LSTM, etc.) to be implemented entirely in hardware.

The invention relates to a method for manufacturing a large hybrid component for an aircraft comprising the following steps: -Production of a basic component (1) of the hybrid large component in its basic form, wherein -an additional functional structure (2) and a guide contour (3) reduced in scale with respect to a target shape of an outer contour (4) of the basic component (12) are applied to the basic component (1) by additive manufacturing, and -a mobile hand tool with a reference surface is provided for producing the desired shape of the outer contour (4), and -the outer contour (4) of the base component (1) is machined by the hand tool by guiding the hand tool with the reference surface on the guide contour (3) during the movement.

This invention relates to the fields of computer network technology and game server technology, specifically to a method and system for automatic scaling up and down and load balancing of distributed game servers. The method involves: configuring operational indicators and corresponding thresholds to trigger scaling up and down; collecting real-time data and comparing it with the thresholds; and determining whether to trigger a scaling up or down process based on the comparison result. If a scaling up process is triggered, server resources are automatically used to deploy new stateful service nodes, synchronize basic data and service registration, and allocate new users to the new stateful service nodes using a load balancing algorithm. If a scaling down process is triggered, stateful service nodes of the combat suit type are scaled down. This invention can achieve fully automated scaling up and down and precise load balancing, improving service response efficiency and optimizing resource utilization.

The invention provides an information processing method, electronic equipment, a storage medium and a computer program product, and the method comprises the steps: monitoring the number of producers, and enabling the producers to be deployed in a container arrangement class framework; when it is determined that the number of producers changes, it is determined that capacity expansion or capacity reduction needs to be conducted on consumers, and the consumers are deployed in the container arrangement class framework; according to the number of the producers and first information, capacity expansion or capacity reduction is carried out on the consumers, and the first information represents the proportional relation between the number of the producers and the number of the consumers. According to the scheme provided by the invention, the universality is good, the business fluctuation can be responded more quickly, and the capacity of consumers can be expanded and shrunk in time.

The application discloses a weed seed identification method and system based on DEIMv2 and a special model architecture for identification, wherein the model architecture comprises: a multi-scale space adapter which extracts spatial features and performs multi-scale down-sampling on a target image to output three spatial feature tensors with descending resolutions; a visual basic model pre-trained in a self-supervised manner which encodes deep semantic features of the target image and outputs multi-scale deep semantic feature tensors; a hybrid encoder which aggregates the spatial feature tensors and the deep semantic feature tensors at the same scale to obtain first, second and third aggregation feature tensors with descending scales; a feature enhancement path is: the first aggregation feature tensor is fused with the second aggregation feature tensor after spatial channel down-sampling, and then fused with the third aggregation feature tensor after spatial channel down-sampling to output an enhanced feature tensor; and a DEIM decoder and a detection head, wherein the decoder receives the enhanced feature tensor and decodes.

The invention provides a high-precision sub-nanometer displacement table. The high-precision sub-nanometer displacement table comprises a guide switching hinge, a displacement zooming driving assembly and a switching frame assembly. The displacement scaling driving assembly comprises a displacement scaling driving hinge and a piezoelectric actuator, the displacement scaling driving hinge is of an integrated sheet structure, the piezoelectric actuator is in contact with the input end of the displacement scaling driving hinge in a magnetic pre-tightening mode, an eccentric hinge structure is arranged in the displacement scaling driving hinge, and the piezoelectric actuator is in contact with the input end of the displacement scaling driving hinge. The displacement scaling driving assembly is used for scaling down the input displacement of the piezoelectric actuator according to the proportion and then outputting the input displacement through the displacement output block, the guide transfer hinge is in hard connection with the displacement output block of the displacement scaling driving assembly, and the transfer frame assembly is used for fixedly installing the guide transfer hinge and the displacement scaling driving assembly. According to the invention, precise positioning requirements under harsh working conditions such as semiconductor manufacturing and synchrotron radiation beam line regulation and control can be met.

The invention discloses a high-resolution unmanned aerial vehicle multi-scale remote sensing image target detection method and device, and belongs to the technical field of target detection, and the method comprises the steps: introducing a target activation network into a backbone network to generate an object activation graph, cutting out a region block containing a target from a feature level through uniform cutting and mapping, and reducing the calculation; a space-channel collaborative attention module and an enhanced multi-level feature reconstruction module are introduced into the neck network to process each layer of feature map and the activation enhancement feature map output by the backbone network to obtain and output high-scale fusion features of each level and low-scale down-sampling features of each level; a dynamic mask sparse calculation module is introduced into the detection network to generate a hard mask of a key area in neck network output features, sparse convolution of the neck network output features is guided by the hard mask to obtain sparse features, classification and bounding box regression are executed after multi-scale sparse feature fusion is performed on all levels of sparse features, and a target detection result is output. The detection speed and the detection accuracy can be improved.

The present application relates to the technical field of computers. Disclosed are an image scaling method and apparatus. The method comprises: receiving an instruction for instructing the scaling-down and scaling-up of an image frame, and calculating the ratio of the height of the adjusted image frame indicated in the instruction to the height of a display screen of a target vehicle; and on the basis of the ratio, scaling down or scaling up an original image frame, which comprises information required to be displayed in a dialog service flow interface, so as to obtain a target image frame, which comprises the information required to be displayed in the dialog service flow interface, and displaying the target image frame in the dialog service flow interface.

A device may be configured to receive a bitstream. The device may decode the bitstream according to a video coding technique. The device may reconstruct feature data from the decoded bitstream. The device may determine a normalization technique from one or more normalization techniques. The device may perform the normalization technique on reconstructed feature data to generate feature data for input into a neural network. Normalizations techniques may include a combination of a linear normalization techniques, scaling down normalization techniques, and logarithmic normalization techniques. In one example, a normalization technique may include a transform based normalization technique. In one example, parameters of a transform based normalization technique may be set based on a picture resolution.

This invention relates to power electronics technology, specifically to a method for transient stability analysis of multi-inverter systems based on a two-layer iterative equal-area method. The method involves establishing a mathematical model of a multi-grid-connected inverter parallel system, defining the equivalent mechanical power, electromagnetic power, other-machine electromagnetic power, electromagnetic interaction power, and self-damping and mutual damping of the system. First, non-integrable terms are scaled down to make them integrable. Then, using the proposed two-layer iterative equal-area method, the upper and lower boundaries of the power angle stability of the multi-grid-connected inverter parallel system are obtained through two iterations. This method, by combining iteration and scaling, improves the conservatism of the stability region estimate as much as possible without causing stability misjudgments. Using the two-layer iterative equal-area method, a relatively accurate estimate of the stability region of the grid-connected inverter parallel system is obtained, improving the conservatism of the original equal-area method applied to VSC parallel systems. It has good development potential and room for promotion.

The present technology provides intercloud connectivity as a service by discovering components of the organization's deployment in various sites, irrespective of the cloud provider, such that two sites can merely be selected along with a few standard options, and the controller can handle the complexity of instantiating a tunnel between the cloud sites automatically. Further, the controller can monitor the health of one or more tunnels between the cloud sites to automatically scale bandwidth up or down.