9results about How to "Generalizable" patented technology

The invention relates to the technical field of intelligent manufacturing and numerical control machining, and particularly discloses a three-dimensional volute model revolution surface turning blank contour line extraction method which comprises the steps that a three-dimensional volute model is input, and all turning revolution surfaces are extracted from the three-dimensional volute model; traversing each turning rotation surface, constructing a plane, and intersecting the plane with the corresponding turning rotation surface to obtain an intersection line corresponding to each turning rotation surface; scanning each intersecting line around a rotating shaft to obtain a corresponding ring surface, and intersecting each ring surface by using the same reference plane to obtain a contour line of each ring surface on the same reference plane; all the contour lines are subjected to duplicate removal and sorting; connecting the sorted contour lines to obtain a closed line frame; and the closed wireframe is swept around the rotating shaft, so that a machined blank body of the three-dimensional volute model is formed. According to the method, various characteristics of the volute revolution surface can be accurately identified, and the method has certain generalization.

The application relates to a kind of mitochondria-based single cell feature extraction and analysis method and system, comprising: obtaining the multiple modal images such as bright field image, nucleus fluorescent image and mitochondria fluorescent image of single cell;Image preprocessing is carried out to the three modal images obtained;For different structures such as mitochondria, morphological and texture features are extracted, and feature analysis is carried out;Further, through the fusion of mitochondria and machine learning technology, the automatic classification of cells is realized.The application is used for the classification of human cervical epithelial cells (H8) and cervical cancer cells (HeLa), and the machine learning analysis of morphological features and texture features shows the potential of mitochondria in the classification of cervical cells.The application has strong applicability, can be combined with machine learning and other analysis methods, and can be applied to various biological cells, has universality, and is easy to popularize.

This application discloses a method, apparatus, device, and medium for detecting loitering events. In the embodiments of this application, it is first determined whether the target object has moved based on the first detection box in the first image and the second detection box in the second image. After determining that the target object has moved, it is further determined whether the target object is moving inside or outside the detection area, thereby judging whether the target object is loitering. This effectively improves the accuracy of loitering event detection and reduces the false alarm rate. Furthermore, this application can determine whether the target object is loitering based on whether it moves back and forth inside or outside the detection area, thus compensating for the blind spots of existing loitering detection algorithms. Specifically, it compensates for the problem that loitering detection algorithms based on dwell time fail when the target object loiteres at the edge of the detection area. Moreover, the technical solution protected by this application has reliability, real-time performance, and generalizability, meeting the characteristics of dependability.

The application provides a self-adaptive adversarial training method for dynamic visual cabinet recognition, comprising the following steps: S1, initializing target network parameters or initializing target network pre-training configuration, obtaining X clean correctly recognized and not disturbed in a dynamic visual cabinet; S2, generating an adaptive adjustment attack parameter vector theta by using a heuristic differential evolution algorithm of a strategy generator according to the robustness of the target network; S3, inputting the attack parameter vector theta into an adversarial sample generator, generating an adversarial sample by adding disturbance in the X clean ; S4, inputting the X clean and the adversarial sample into the target network for training, and setting a training target function; S5, repeating steps S2-S4 until the maximum iteration number is reached, and obtaining a recognition model for dynamic visual cabinet recognition. The application can improve the adversarial attack capability of the recognition model in the dynamic visual cabinet.

This invention relates to a remote sensing image scene classification method based on uncertain visual-language alignment, comprising the following steps: acquiring remote sensing images as training and testing sets; using a CLIP dual encoder to map the original image and text to a shared point representation space; the uncertain encoder predicting each embedding vector along two paths; internally, the uncertain encoder performing two-level interactions on the input embedding vectors; globally modeling the image and text as two Gaussian distributions based on the mean and variance vectors, maximizing the image-text similarity that should be correctly matched in the entire batch; acquiring new remote sensing images and inputting them into the trained model, which then classifies the scenes in the remote sensing images. This application captures fuzzy mappings between modalities in the probability space, providing a more generalized representation for downstream remote sensing image scene classification tasks.

This invention belongs to the field of radar information processing technology and discloses a pre-training method for a multimodal model of radar target detection, tracking, and recognition. The invention includes collecting model pre-training data; converting the collected radar information processing data from various models into a unified data format through data preprocessing to serve as input data for the pre-trained model; achieving simultaneous input of multimodal data, including images, tracks, and category text, through a unified pre-training framework, and completing joint representation learning of the input data by the pre-training framework; and designing a multi-task joint learning optimization algorithm to achieve parallel training of radar target detection, tracking, and recognition tasks. This invention achieves integrated processing of radar target detection, tracking, and recognition tasks by establishing a multi-task learning mechanism under a unified model training framework, effectively improving the overall performance of radar information processing. Simultaneously, by enhancing versatility and generalization, the model is applicable to intelligent processing tasks of various radar models.

This invention discloses a method and apparatus for detoxifying large language models based on Euclidean flow matching, belonging to the field of artificial intelligence. To better balance detoxification effect and generation quality, this invention extracts features from the LLM output using a sparse autoencoder, performs detoxification transformation based on Euclidean flow matching, and finally replaces the original input activation, guiding the LLM to generate non-toxic content. This invention, applied to the field of large language models, can efficiently detoxify LLMs without significantly sacrificing fluency.

The present application belongs to the technical field of intelligent video processing, and particularly relates to a future frame anomaly detection method based on meta learning and space-time relationship. The present application method proposes a meta learning module, and a model-based meta learning method enables the model to learn general features from multiple tasks, so that the system can learn more discriminative and generalizable feature representations from data. In different monitoring scenes and abnormal behaviors, the model can also accurately perform anomaly detection without overfitting or underfitting. The present application method introduces the meta learning module into the autoencoder, uses the learning characteristics of the meta learning module to extract, save and update the features extracted by the autoencoder, automatically acquires the feature importance of the input video frame through learning, and assigns important weights to the features that are more worthy of attention, which helps to improve the utilization rate of key features in the input stage of the future frame prediction network.

The invention provides a gene regulatory network pathway efficiency characterization method and system based on methylation modification, and the method comprises the steps: firstly carrying out the structural preprocessing of original multi-omics data, including a gene expression matrix and a DNA methylation matrix, and carrying out the feature screening, circuit-type topology reconstruction and signal flow mapping, thereby obtaining the performance of a gene regulatory network pathway; generating directed gene regulation networks in one-to-one correspondence with the samples; and on the basis, extracting a gene subset corresponding to a known pathway in the network, constructing a strongly connected pathway subgraph, further calculating the steady-state transfer information amount of the network by utilizing Markov chain entropy, and forming a quantitative characterization index of a system function so as to screen key pathway information. The network hierarchy provided by the scheme is more detailed, the modes of transcriptional activity inhibition and reactivation can be identified, and the biological interpretation is better; the order quantitative index provided by the scheme is more significant, the algorithm is concise, and the method has better expansibility and generalization.