4 results about "Class activation mapping" patented technology

The application discloses a method and system for predicting protein drug binding sites based on a multi-modal dynamic graph, comprising: obtaining amino acid sequence data and three-dimensional structure data of a protein, and generating evolutionary conservation features, structure graph topology features and sequence features respectively; inputting the features into a multi-modal fusion encoder to generate first fusion features; inputting the first fusion features into a prediction decoder to generate initial prediction probabilities; iteratively updating the structure graph topology features according to a preset three-graph update rule according to the initial prediction probabilities and residue dynamic communication scores; re-inputting the updated structure graph topology features into the multi-modal fusion encoder and the prediction decoder, and repeatedly executing until a preset iteration number is reached, to generate final prediction probabilities; and generating a residue importance heat map and outputting a prediction report based on the final prediction probabilities through a gradient weighted class activation mapping algorithm. The application realizes the cooperative optimization of prediction and graph structure, and significantly improves the accuracy of binding site prediction.

The application discloses an agricultural seed screening method and system based on data analysis, relates to the technical field of data processing, obtains a transmission mode hyperspectral image of a seed to be tested, obtains an optimal contrast reference image based on an OTSU method and generates a global binary mask, obtains an effective connected region based on the global binary mask, performs full-waveband data cutting on the transmission mode hyperspectral image, generates a region of interest, determines a key characteristic wavelength through a one-dimensional deep full convolution neural network model and a class activation mapping algorithm, calculates geometric morphological feature data of the seed to be tested in the region of interest, obtains an optimal spectral feature vector of the seed to be tested based on the key characteristic wavelength, constructs an original geometric feature vector and an original spectral feature vector, performs normalization processing, generates a graph fusion feature vector, constructs a dynamic classification model, and if it is determined that the seed to be tested falls into an unqualified area or a risk buffer area, drives a pneumatic nozzle to realize physical separation.

This invention relates to the field of medical image artificial intelligence analysis technology, specifically to an automated preoperative precise risk stratification system and method for gastrointestinal stromal tumors. It includes: a data preprocessing module for acquiring the original sequences of multi-phase computed tomography (CT) scans of the patient's abdomen, and performing three-dimensional volume data construction, spatial dimension standardization, and voxel intensity normalization to generate a standardized three-dimensional tensor; a deep learning model module for receiving the three-dimensional tensor and, based on a 3D Swin Transformer architecture, outputting classification probabilities corresponding to four risk levels: very low risk, low risk, intermediate risk, and high risk; and an interpretable visualization module for generating heatmaps identifying key decision-making regions of the model using gradient-weighted class activation mapping technology, and overlaying these heatmaps back onto the original CT images for visualization.

The present application belongs to the technical field of fire prediction, and relates to a parking lot fire prediction method. The method positions a fire area through class activation mapping based on semantic features, and uses a cross-generation feature aggregation module to aggregate iterative output results based on network intermediate layer multi-scale features and results output during a training process. Meanwhile, the method fuses an original RGB input picture based on an initial rough segmentation result, and inputs the picture into a classification network for secondary fine segmentation, which can reduce a false positive rate, is not affected by a surrounding environment, has low data requirements, and can only be used in the field of parking lot fire detection, and is also applicable to other fields with insufficient data sets and requiring a weak supervision mode to improve precision.