Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

2450results about How to "Easy alignment" patented technology

Audio generation method and device, electronic equipment and storage medium

The embodiment of the invention discloses an audio generation method and device, electronic equipment and a storage medium, and the method comprises the steps: extracting first text features of an object text used for describing a target object, and carrying out the weighted summation of the first text features in a same video frame image through a mask score, according to the method, the second text feature representing the target object in the video frame image more accurately can be obtained, and on the basis, the target text feature is obtained by fusing the second text features of the plurality of video frame images, so that the target text feature can provide frame-level and fine-grained semantic information; and subsequently, obtaining a randomly generated audio signal with noise, denoising the audio signal with noise by taking the target video feature and the target text feature as denoising conditions to obtain a target audio signal, and generating a target audio corresponding to the target video based on the target audio signal to realize frame-level audio generation. The alignment effect of the pictures in the target audio and the target video is optimized, so that the audio generation quality is improved.
Owner:TENCENT TECHNOLOGY (SHENZHEN) CO LTD

Fabricated building quality monitoring system based on BIM and point cloud integration

The invention provides a fabricated building quality monitoring system based on BIM and point cloud integration, and relates to the technical field of data processing, and the system comprises a BIM model feature point obtaining module which is used for obtaining BIM model feature points; the point cloud feature point obtaining module is used for obtaining point cloud feature points; the preliminary matching result obtaining module is used for carrying out descriptor matching and matching pair screening; the optimization matching result obtaining module is used for performing evaluation transformation and error calculation; the fine alignment result obtaining module is used for carrying out preliminary transformation calculation, nearest neighbor searching and error minimization; and the quality improvement suggestion generation module is used for carrying out error analysis and manual correction and carrying out monitoring and quality tracing. According to the method and the device, the technical problem of insufficient feature point extraction and matching precision in the prior art can be solved, and the technical effect of improving the alignment precision of the point cloud data and the BIM model by optimizing the feature point extraction and matching precision is achieved.
Owner:SHANDONG JIANZHU UNIV +1

Diffusion model LoRA fine tuning optimization method and system based on CLIP loss and perception loss

The invention provides a diffusion model LoRA fine tuning optimization method and system based on CLIP loss and perception loss, and the method comprises the steps: 1, dynamically adjusting the weights of the CLIP loss and the perception loss in combination with the CLIP loss and the perception loss in a LoRA fine tuning process; 2, calculating the semantic similarity between the denoised intermediate image and the target text by using a CLIP model, and optimizing the noise prediction capability of a diffusion model according to the similarity difference; step 3, calculating the difference between the intermediate image and the target image in the feature space by using perception loss, optimizing the noise prediction capability of the diffusion model, and improving the visual quality and detail fidelity of the generated image; and step 4, adjusting whether to start CLIP loss and perception loss or not according to the training progress. By introducing the CLIP loss, the model can better align the image and the text in the fine tuning training process, so that the generated image better conforms to the description of the text prompt.
Owner:NANJING UNIV OF INFORMATION SCI & TECH

Zero sample image classification method and device

The invention discloses a zero sample image classification method and device, and belongs to the field of image recognition. According to the method, a category description document is generated through a large language model, global and local feature information of an image and the document is obtained through a visual encoder and a text encoder, then decoupling is performed through a visual semantic decomposition module and a text semantic decomposition module, and visual semantic embedding and text semantic embedding are generated; through joint training of the modules, optimizing total loss composed of local-to-semantic mapping variance loss, multi-semantic diversity loss, decomposition semantic alignment loss and local semantic alignment loss; and the training is used for image category prediction in the reasoning stage after completion of training. According to the invention, the difficulty of zero sample image classification can be reduced, and the classification accuracy and efficiency are improved.
Owner:INSTITUTE OF INFORMATION ENGINEERING CHINESE ACADEMY OF SCIENCES

Solid-state battery cell manufacturing method and device based on single-sided rubber frame molding

The invention discloses a solid-state battery cell manufacturing method and device based on single-sided rubber frame forming, and relates to the technical field of solid-state battery manufacturing. The method comprises the following steps: manufacturing a rubber frame on one surface of a first pole piece positioned at a forming station; the second pole pieces and the first pole pieces with the rubber frames are sequentially and alternately conveyed to a lamination station, and the first pole pieces and the second pole pieces are sequentially and alternately laminated, so that the second pole pieces are embedded into the rubber frames of the first pole pieces; wherein the electrodes of the first pole piece and the second pole piece are opposite, and the thickness of the rubber frame is greater than or equal to that of the second pole piece. According to the invention, two times of rubber frame forming on one pole piece are not needed, so that the connectivity between the pole pieces is improved, and the conditions that the solid electrolyte layer and the pole pieces deviate relatively and the two adjacent pole pieces are short-circuited are avoided.
Owner:GUANGDONG LYRIC ROBOT INTELLIGENT AUTOMATION CO LTD

Long sequence action recognition method for small sample learning

The invention provides a long sequence action recognition method for small sample learning, and aims to solve the problem of high calculation complexity of a traditional Transform-based method during long sequence processing, improve the capabilities of local feature modeling and time sequence alignment and relieve the negative influence of intra-class difference accumulation on model performance. The method mainly comprises three core parts: a feature extraction module, a Mamba branch module and a comparative learning module. The invention provides a multi-scale Matryoshka Mama model, aiming at the limitation of a traditional Mama method in small sample action recognition. According to the model, the expression ability of local features is enhanced through the inner module, and accurate alignment of the local features is realized through the outer module. In addition, the invention further provides a mixed contrast learning method which is used for effectively relieving the negative influence of intra-class difference accumulation of long sequences of the same class on small sample action recognition performance. According to the method, the accuracy and robustness of long sequence action recognition are remarkably improved while the small sample learning efficiency is improved.
Owner:SOUTHEAST UNIV

Automatic air-cooled reactor loading line for hydrogen fuel cell stack production

The invention relates to the technical field of hydrogen fuel cells, in particular to an air-cooled reactor automatic loading line for hydrogen fuel cell stack production, which comprises a rack, a storage rack, a material moving mechanism and two conveyor belts, the material moving mechanism comprises a support frame, a driving assembly and a pressing assembly, and electrode plates and bipolar plates are alternately conveyed to the material moving mechanism through the two conveyor belts. The electrode plate or the bipolar plate is supported through the sliding rollers at the bottoms of the overturning frames, and then the two overturning frames are synchronously driven to move through the driving assembly, so that the electrode plate or the bipolar plate naturally falls between the two overturning frames. The auxiliary electrode plate or the bipolar plate can stably fall on the storage rack in cooperation with downward pressing of the downward pressing assembly; the process does not need manual intervention, and the accuracy and regularity of stacking are ensured. The automatic stacking reduces the dependence on manual operation, reduces the labor cost, and further reduces the rework cost caused by unqualified materials due to the reduction of the surface damage of the electrode plates and the bipolar plates.
Owner:XIE HYDROGEN (SHANGHAI) NEW ENERGY TECH CO LTD

Medical image classification method based on prior knowledge enhanced mask and alignment modeling

The invention belongs to the technical field of computer vision, and particularly relates to a medical image classification method based on priori knowledge enhanced masks and alignment modeling. Comprising the following steps: acquiring an image text pair data set and preprocessing the image text pair data set to obtain a preprocessed image text pair data set; performing preliminary training on a visual encoder and a text encoder according to the preprocessed image text pair data set to obtain a preliminarily trained visual encoder and a preliminarily trained text encoder; training the visual encoder for the second time in combination with the text encoder to obtain a trained visual encoder; classifying the medical images by adopting the trained visual encoder; the model provided by the invention can be better generalized to various downstream medical image recognition tasks, and the accuracy of the downstream medical image recognition tasks is improved.
Owner:CHONGQING UNIV OF POSTS & TELECOMM

Cross-modal foreground distillation method for multi-view three-dimensional target detection

The invention discloses a cross-modal foreground distillation method for multi-view three-dimensional target detection, and belongs to the technical field of automatic driving perception. According to the method, a laser radar or a multi-modal three-dimensional target detection model is used as a teacher model, a multi-view aerial view BEV perception three-dimensional target detection model is used as a student model, and the model performance is improved through multi-scale feature distillation; a space attention mechanism and a foreground mask module are introduced, and the student model can simulate feature responses of the teacher model in different space positions under multiple scales, focus on a foreground target and suppress background noise; the adaptive module ensures the consistency of the multi-scale features in feature dimensions and spatial resolution, and finally, the difference between the features and regression output is measured through loss functions such as mean square error and KL divergence, so that efficient cross-modal and multi-scale foreground feature distillation is realized. Finally, the multi-view three-dimensional target detection precision and efficiency are improved, and meanwhile, the distillation requirements of multi-scale features are fully considered.
Owner:KUNMING UNIV OF SCI & TECH

Multi-modal emotion prompting method and system for online auxiliary voice dialogue scene

The invention discloses a multi-modal emotion prompting method and system for an online auxiliary voice dialogue scene, and relates to the technical field of artificial intelligence, voice signals in the voice dialogue scene are input into a trained multi-modal emotion prompting large model, and emotion labels and voice emotion prompting information are output; the training process of the multi-modal emotion prompt large model comprises the following steps: acquiring original voice data of a real dialogue scene to construct a training data set for emotion recognition; performing feature extraction on the original voice data in the training data set to obtain an original voice feature vector; inputting the original voice feature vector into a sparse bridging Transform module, and outputting a fusion feature; inputting the fusion features into a large language model, generating coherent voice emotion prompt information through text decoding, and associating with corresponding emotion tags; according to the multi-mode emotion prompting method and system, accurate analysis and prompting of voice emotion in an online auxiliary voice dialogue scene are realized.
Owner:ARTIFICIAL INTELLIGENCE RES INST OF HEFEI COMPREHENSIVE NAT SCI CENT (ANHUI ARTIFICIAL INTELLIGENCE LAB)

Solar cell

The back surface of the solar cell is provided with metal grid lines located in a grid line area and alignment marks located in a non-grid line area. The solar cell further comprises a patterned passivation contact structure, the passivation contact structure only comprises a first area located in the grid line area and a second area located at the position where the alignment marks are formed, the metal grid lines are located in the first area, and the alignment marks are located in the second area. According to the utility model, the alignment mark is arranged on the second area of the non-grid line area, so that the contrast around the alignment mark can be increased, the point grabbing precision is improved, and good grid line alignment can be realized.
Owner:扬州阿特斯太阳能电池有限公司 +1

Automatic flange centering and welding device

The utility model relates to the technical field of flange centering welding, in particular to an automatic flange centering welding device which comprises a base, a centering mechanism installed on the base, a pressing air cylinder installed on the base and used for pressing a flange and a displacement mechanism used for driving the base to move in a reciprocating mode. The centering mechanism comprises a centering assembly and a power part, the power part drives the centering assembly to synchronously move towards the periphery in the radial direction and is fixedly installed relative to the base, the centering assembly synchronously moves towards the periphery in the radial direction and is used for supporting the inner diameter of a steel pipe, and a positioning pin is further installed on the centering mechanism. The positioning pin is used for solving the technical problems that in the prior art, when a steel pipe and a flange are subjected to centering welding, the steel pipe is prone to deformation, and the product rejection rate is high due to the poor centering performance of the steel pipe and the flange.
Owner:HAODE BOER (SHANDONG) INTELLIGENT EQUIP CO LTD

Solar frame and lamp pole welding device and method based on solar street lamp processing

The invention relates to the technical field of welding devices, and particularly discloses a solar frame and lamp pole welding device and method based on solar street lamp machining. Comprising a fixing base, a monitoring mechanism arranged at the top of the fixing base and used for determining the attaching degree of a lamp pole sleeve and a frame body, and a limiting mechanism arranged in linkage with the monitoring mechanism and used for fixing a solar frame. Therefore, the abutting degree of the lamp pole sleeve and the solar frame can be judged conveniently, the situation that welding is unstable due to a gap between the lamp pole sleeve and the solar frame is reduced, the limiting assembly and the monitoring assembly are arranged in a linkage mode, the solar frame can be clamped conveniently, the inner wall of the lamp pole sleeve can be aligned to the center of the solar frame conveniently, and the working efficiency is improved. And therefore, the outer wall and the inner wall of the lamp pole sleeve can be welded before and after subsequent overturning, and the welding stability of the lamp pole sleeve and the solar frame is improved.
Owner:胡云秋

Blasting drilling device for tunnel precise blasting construction and using method of blasting drilling device

The invention relates to the technical field of drilling equipment, in particular to a blasting drilling device for tunnel precise blasting construction and a using method thereof.The blasting drilling device comprises a mounting main body, a rotating mechanism is arranged on the mounting main body, a connecting assembly is arranged on the rotating mechanism, and a translation mechanism is arranged on the connecting assembly; a limiting assembly is further arranged on the mounting body, a limiting adjusting mechanism is arranged between the translation mechanism and the mounting body, and a drilling machine is fixedly connected to the translation mechanism. According to the drilling machine, through the arranged limiting adjusting assembly, in the process that the sliding plate drives the drilling machine to drill and complete resetting, a rack can be driven to move and be engaged with a gear ring, the gear ring is made to rotate, accordingly, a limiting stop lever is driven to rotate by a certain angle, and the rotated limiting stop lever serves as a position mark and can be used for controlling the rotating angle of the drilling machine; and accurate positioning of the drilling machine is ensured.
Owner:JILIN LONGYE BLASTING ENG CO LTD

Piece-by-piece feeding device for box body assembly and method of piece-by-piece feeding device

The invention discloses a piece-by-piece feeding device and method for box body assembling, and belongs to the technical field of box body machining. The piece-by-piece feeding device comprises a rack, and two pushing assemblies used for preliminarily feeding stacked thick paperboards and aligning the stacked thick paperboards in the left-right direction are symmetrically installed on the leftmost side of the rack in the front-back direction; the right side of the pushing assembly is provided with a feeding assembly used for aligning the stacked thick paperboards front and back and alternately feeding the stacked thick paperboards in a double-station mode, the middle side of the feeding assembly is provided with a conveying assembly used for feeding the stacked thick paperboards one by one, and the right side of the feeding assembly is provided with a board supporting assembly used for assisting in supporting the thick paperboards for feeding. And a gluing machine for gluing the thick paperboard is arranged on the right side of the supporting plate assembly. By means of the mode, sheet alignment in the front-back direction and the left-right direction can be carried out while double-station alternate feeding is carried out, continuous sheet-by-sheet feeding can be carried out on sheets, and machining of the next station is facilitated.
Owner:HEFEI HUAGUAN PRINTING

Consistency identity picture generation method based on diffusion model

The invention belongs to the field of computer vision, and particularly relates to a consistent identity picture generation method based on a diffusion model, which comprises the following steps of: generating optimal paired data by adopting a multi-modal large language model, the paired data comprising text data and corresponding image data; carrying out encoding processing on the text data by adopting a text encoder to obtain a text embedding vector; processing the image data by adopting an image encoder to obtain an image embedding vector; inputting the text embedding vector and the image embedding vector into an MLP module for fusion to obtain a fusion embedding vector; stacking all the fusion embedding vectors, and processing the stacked fusion embedding vectors by adopting a cross attention layer to obtain deep-level features; performing multi-stage adjustment on the deep feature to obtain a consistent identity picture; according to the invention, a coarse-to-fine multi-stage fine adjustment method is adopted, and the visual quality of the generated image is gradually optimized.
Owner:SHANGHAI-CHONGQING ARTIFICIAL INTELLIGENCE RES INST

Multi-layer stacked chip packaging method

The embodiment of the invention provides a multi-layer stacked chip packaging method, which comprises the following steps of: forming a plurality of chips, and forming conductive bumps on the first surfaces of the chips; forming a first plastic package body wrapping the chip and forming a first interconnection layer on the exposed conductive bumps to form a first chip module; forming a redistribution layer, forming a first convex column on the redistribution layer, and forming a dielectric column on the first convex column; fixing a chip on the rewiring layer and forming a second plastic package body on the rewiring layer; removing the dielectric columns to form blind holes, and forming second convex columns on the first convex columns to form plastic package body conductive columns; forming a second interconnection layer on the surface, deviating from the rewiring layer, of the second plastic package body to form a second chip module; and sequentially stacking the plurality of second chip modules on the substrate, and stacking the first chip module to the second chip module on the uppermost layer. According to the method, the structure storage capacity is improved, and the alignment precision requirement is reduced; the surface uniformity of the first convex column is improved when the second convex column is formed, and the stability and the reliability of the conductive column of the plastic package body are improved.
Owner:NANTONG FUJITSU MICROELECTRONICS

Pipe welding device for building construction

The invention belongs to the technical field of building construction equipment, and particularly relates to a pipe welding device for building construction, which comprises a supporting bottom plate, a supporting main beam is fixedly connected to the supporting bottom plate, a transverse mounting beam is fixedly connected to the supporting main beam, and two extension plates are fixedly connected to one side of the transverse mounting beam. By arranging the pipeline clamping movement mechanism, the two vertical platens can get close to each other or get away from each other, so that movement of two sections of pipelines is achieved, and by arranging a threaded rod and a second nut, movement of a fixed roller can be controlled, so that clamping work of pipelines of different sizes is achieved in cooperation with a supporting roller; and a rotating base, a first torsion spring and an arc-shaped scraper are further arranged, when the pipeline is welded, the arc-shaped scraper can be tightly attached to the surface of a welding face, redundant welding slag on the surface of the welding face is removed in time, a matched welding mechanism is further arranged, automation of bar conveying work can be achieved, and welding work is more stable and efficient.
Owner:郑权

High-temperature superconducting Tokamak cold shield assembly assembling system and method

The invention discloses a high-temperature superconducting Tokamak cold shield assembly assembling system and method. The high-temperature superconducting Tokamak cold shield assembly assembling system comprises a positioning assembly, a welding assembly, a supporting frame, supporting legs, a locking mechanism and a driving mechanism. The invention relates to the technical field of welding tools. According to the high-temperature superconducting Tokamak cold shield assembly assembling system and method, the supporting frame is arranged to be installed around the lower-layer high-temperature superconducting Tokamak cold shield, the positioning assembly and the welding assembly can be placed on the supporting frame in sequence, and the positioning assembly can abut against the lower-layer high-temperature superconducting Tokamak cold shield by half; the high-temperature superconducting Tokamak cold shield suspended on the upper layer to be put down is positioned and guided, it is guaranteed that the high-temperature superconducting Tokamak cold shield can be accurately in butt joint with the high-temperature superconducting Tokamak cold shield on the lower layer, and then the problem that butt joint dislocation of the heavy high-temperature superconducting Tokamak cold shield is difficult to adjust is solved. And after the positioning assembly is taken down, the welding assembly is replaced, the welding assembly can be driven by the driving mechanism to rotate around the high-temperature superconducting Tokamak cold shield along the supporting frame for welding, and manpower is saved.
Owner:无锡华立聚能装备股份有限公司

Solid elastic card heat regeneration device easy to assemble

The invention provides an easy-to-assemble backheating device for a solid elastic card, which comprises a sleeve, a plurality of solid elastic card material plates stacked in the sleeve and a pressure head for applying stress to the solid elastic card material plates, and the sleeve is composed of a plurality of parts and is used for inspecting or assembling and disassembling the plurality of solid elastic card material plates when being unfolded; each solid elastic clamping material plate is provided with a flow cavity after being stacked, a heat-conducting medium exchanges heat with the solid elastic clamping material plates through the flow cavities when heat or cold is generated, and the heat-conducting medium flows into a heat exchanger through a liquid through hole in the pressure head for heat exchange. The sleeve is designed to be a plurality of detachable parts or folding parts, during assembly, a plurality of solid elastic clamping material plates can be conveniently placed in, alignment is convenient, in the loading and unloading process, even if a certain solid elastic clamping material plate goes wrong, opening and inspection are convenient, and therefore the assembly efficiency is improved, and the fault solving efficiency is improved.
Owner:SHENZHEN ENTROPLUS INNOVATION TECHNOLOGY CO LTD

Image-text retrieval method based on information enhancement and multi-modal global local feature alignment

The invention discloses an image-text retrieval method based on information enhancement and multi-modal global local feature alignment, and relates to the technical field of image-text retrieval, and the method comprises the following steps: 1, obtaining image information and corresponding text description data; 2, respectively carrying out feature extraction on the picture data and the text data to obtain picture global features, local features, text word features and sentence features; 3, fusing the picture local features and the picture global features to obtain local features with global information, and realizing information enhancement; 4, putting the image features and the text features into a cross attention mechanism model for coarse and fine granularity feature fusion to obtain image-text global fusion features and image-text local fusion features, and realizing coarse and fine granularity alignment of the image-text; and 5, carrying out similarity calculation on the image-text local fusion feature and the image-text local feature to realize image-text retrieval. Through information enhancement and feature fusion, alignment of coarse granularity and fine granularity is ensured, and alignment of images and texts is improved.
Owner:SOUTHWEST JIAOTONG UNIV

Device for installing insulating sheath of transformer pile head in live-line mode

The invention relates to the field of insulating sheath installation, in particular to an electrified installation transformer pile head insulating sheath installation device which comprises a first installation rod, a second installation rod and a connecting rod, an insulating sheath is installed in the installation rod, a wire sheath is installed on the insulating sheath, and wire grooves are formed in the same horizontal line of the insulating sheath and the wire sheath. A linear motor is installed in the installation sleeve, the output end of the linear motor is connected with the top of the insulation sheath, an empty groove is formed in the installation sleeve, a first spring is installed in the empty groove and connected with a clamping block, a top block is installed on the top of the insulation sheath, and the side, close to the clamping block, of the top block is provided with an inclined face matched with the side face of the clamping block. A first mounting disc is mounted on the outer side of the mounting sleeve, a driving structure is arranged in the first mounting disc, and the driving structure is used for driving the mounting sleeve to rotate so as to adjust the mounting direction of the insulating sheath and the wire sheath, manual mounting is not needed, the risk of safety accidents is reduced, and the life safety of operators is guaranteed.
Owner:STATE GRID ANHUI ELECTRIC POWER CO LTD SHOU COUNTY POWER SUPPLY CO

Packet arranging method, device and equipment for three-dimensional model, medium and product

The invention discloses a three-dimensional model package arranging method, device and equipment, a medium and a product, and relates to the technical field of three-dimensional (3D) printing.The method comprises the steps that space parameters of a to-be-printed model and a printing area are determined, the to-be-printed model comprises i three-dimensional models, and i is a positive integer larger than 1; based on the model parameters of the i three-dimensional models, the arrangement sequence of the i three-dimensional models is determined, and the model parameters are the contour parameters of the minimum enclosing rectangles of the three-dimensional models; and the i three-dimensional models are placed in the printing area based on the space parameters of the printing area and the placing sequence of the i three-dimensional models by utilizing a preset model printing rule. According to the method, the arrangement uniformity of the printing models can be improved while the package arrangement speed is ensured, the to-be-printed models are arranged in the printing area as many as possible, excessive analysis on the to-be-printed models is avoided, the calculation amount of a processor is saved, package arrangement efficiency and package arrangement quality are both considered, and the package arrangement experience of a user is improved.
Owner:XIAMEN HANIN CO LTD

Resin tank assembly and water softening device

The utility model relates to the field of water softening, and provides a resin tank assembly and a water softening device.The resin tank assembly comprises a tank body, a waterway component, a central pipe and an adapter, and the tank body is provided with a containing cavity and a tank opening communicated with the containing cavity; the water path component is installed at the end, close to the tank opening, of the tank body and provided with a soft water channel, an adapter communicated with the soft water channel is arranged on the side, facing the tank opening, of the water path component, and the adapter is open; the central pipe is arranged in the containing cavity, and the first end of the central pipe penetrates out of the tank opening; the adapter is connected to the first end of the central pipe and covers the opening of the adapter to form a water guide cavity, and the water guide cavity is communicated with the central pipe and the soft water channel. Connection between the waterway component and the central pipe is simplified, alignment is convenient, and the installation efficiency is improved.
Owner:FOSHAN SHUNDE MIDEA WATER DISPENSER MFG +1

Automatic light beam centering method and system based on laser cutting, laser cutting machine and storage medium

The invention provides an automatic light beam centering method and system based on laser cutting, a laser cutting machine and a storage medium, and belongs to the technical field of laser cutting. The laser cutting head is controlled to move above the light source; turning on a light source, shooting a nozzle image, and analyzing the nozzle image to obtain the center and radius of the nozzle; executing a laser cutting process, acquiring a molten pool center position in real time, and monitoring a relative position of the molten pool center and a nozzle center; if the offset distance between the center of the molten pool and the center of the nozzle is larger than a preset offset threshold value, it is judged that the center of the nozzle is not centered and recorded in a storage, and a lens is controlled to move to achieve centering. Automatic centering can be achieved, the laser centering efficiency is improved, and errors caused by manual operation are reduced. And when deviation occurs, automatic adjustment and centering are performed in real time, so that the consistency of cutting quality is ensured, and the phenomena of dregs, burrs and the like of parts during cutting are effectively prevented.
Owner:JINAN BODOR LASER CO LTD

Cross-modal model training method and device, electronic equipment and storage medium

The invention provides a cross-modal model training method and device, electronic equipment and a storage medium. The method comprises the following steps: based on sample protein data and sample annotation data, calling a cross-modal model to perform feature extraction processing to obtain a first protein feature of the sample protein data and a first text feature of the sample annotation data; encoding the first protein feature and the first text feature to obtain a second protein feature and a second text feature; determining a first sub-comparison loss based on the first protein feature and the second text feature, and determining a second sub-comparison loss based on the second protein feature and the first text feature; determining a total loss function corresponding to the cross-modal model based on the first sub-comparison loss and the second sub-comparison loss; and performing parameter updating on the cross-modal model based on the total loss function to obtain a trained cross-modal model. According to the method and the device, the accuracy of feature extraction of the protein and text cross-modal model can be improved.
Owner:TENCENT TECHNOLOGY (SHENZHEN) CO LTD

Video defogging method and system based on enhanced alignment

The invention belongs to the technical field of computer vision and image processing, and particularly discloses a video defogging method and system based on enhanced alignment, and the method comprises the steps: carrying out the color mapping of a plurality of foggy images through a color lookup table, and obtaining a plurality of enhanced images; based on the enhanced image corresponding to the foggy target frame and the enhanced image corresponding to each foggy reference frame, performing space-time optical flow estimation to obtain a three-dimensional optical flow; performing image transformation on the foggy reference frame based on the three-dimensional optical flow to obtain an alignment frame; and based on the alignment frame, the foggy target frame and the enhanced image corresponding to the foggy target frame, performing splicing and feature fusion to obtain a defogged image corresponding to the foggy target frame. According to the method, details of the foggy image are enhanced, a more accurate alignment result is obtained, more inter-frame information is provided through accurate alignment frames, the network can be effectively helped to learn a better defogging capability, the utilization capability of the deep neural network on the inter-frame information can be improved, and a better video defogging effect is achieved.
Owner:HUAZHONG UNIV OF SCI & TECH

Data processing method and device and model optimization method and device

The embodiment of the invention provides a data processing method and device and a model optimization method and device, and the data processing method comprises the steps: carrying out the sampling in a plurality of data sets, and determining target sample data according to a sampling result; inputting the target sample data into a plurality of language models for processing respectively, and obtaining target prediction data output by the plurality of language models respectively; performing text annotation processing on each piece of target prediction data by using a preset text annotation model to obtain text annotation information corresponding to each piece of target prediction data; based on the target sample data, the target prediction data and the text annotation information, a sample data set is constructed, and the sample data set is used for executing a model training task and a model verification task.
Owner:INFLY TECH (SHANGHAI) CO LTD

Insulator surface laser cleaning device

An insulator surface laser cleaning device comprises a laser cleaning head, a cleaning main machine, a control box, a cleaning robot and a cleaning rotary table, an insulator is vertically connected to the upper side of the cleaning rotary table, the output end of the laser cleaning head is arranged relative to the surface of the insulator, and the output end of the cleaning robot is connected to the laser cleaning head. The control box is connected to the cleaning robot, the cleaning main machine is connected to the laser cleaning head, the cleaning rotary table is used for driving an insulator to rotate along the axis of the cleaning rotary table, the cleaning main machine is used for controlling a switch of the laser cleaning head, and the control box is used for controlling the cleaning robot to drive the laser cleaning head to move. According to the design, the cleaning rotary table drives the insulator to rotate along the axis of the insulator, in the rotating process, the laser cleaning head is aligned with the surface of the insulator through the cleaning robot and the cleaning main machine for laser cleaning, the cleaning efficiency is high, and sweeping is relatively uniform.
Owner:SUZHOU XIANGMING LASER TECH CO LTD

Wind power tower tube welding device and method

The invention discloses a wind power tower tube welding device and method, and relates to the technical field of welding devices. The wind power tower tube welding device comprises a bottom plate, the top of the bottom plate is connected with multiple sets of supporting blocks arranged in an array mode through a first telescopic mechanism, the number of the supporting blocks is two, first balls are arranged at the tops of the supporting blocks, and the supporting blocks are pushed to move through a pushing mechanism; a first supporting plate is fixedly connected to the top of the bottom plate, a first disc is fixedly connected to the upper end of the first supporting plate, and second balls are arranged at the end of the first disc. According to the wind power tower tube welding device and method, adjustment can be conveniently conducted according to the taper of the tower tube, supporting and positioning of the tower tube are more convenient, the effect is better, then the follow-up welding efficiency and quality are guaranteed, before welding, the welding position is conveniently detected, whether defects exist or not is determined, and the follow-up welding quality is guaranteed.
Owner:华能(临高)新能源有限公司 +1