Partial Feature-Set Clustering for Event Data Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing method for selecting frames to be provided with ground truth labels in long movies incurs high calculation costs due to frame-based clustering, which is inefficient for generating training data in machine learning.
Innovation Solution
A method involving dividing feature series into partial feature sets, clustering these sets into groups, and selecting target data with ground truth labels from within these clusters to reduce calculation costs and enhance efficiency in generating training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If frame-based clustering is performed on all frames in long movies, then comprehensive frame classification is achieved, but calculation cost increases significantly
Solution Approach 1:
The patent divides the feature series into multiple partial feature sets (e.g., first partial feature set, second partial feature set) and performs clustering on these segmented sets rather than on all frames simultaneously. This segmentation reduces the computational burden while maintaining classification effectiveness.
Solution Approach 2:
The patent extracts and processes only the necessary partial feature sets for clustering, rather than processing all frame features. By taking out only the essential feature portions needed for classification, the system reduces calculation cost while preserving measurement precision.
2Quantity of substance
If all frames are provided with ground truth labels, then complete training data is obtained, but data processing time and cost increase
Solution Approach 1:
The patent applies partial action by providing ground truth labels only to selected frames rather than all frames. The selection process identifies representative frames from clustered partial feature sets, providing labels to a subset of data that is sufficient for training while significantly reducing processing time and cost.
3Measurement precision
If clustering is performed on per frame basis in long movies, then detailed frame-level analysis is achieved, but operation efficiency decreases
Solution Approach 1:
The patent segments the feature series into partial feature sets before clustering, which allows for detailed frame-level analysis of the segmented portions while improving overall processing efficiency. The segmentation enables parallel processing and reduces the complexity of frame-level analysis.
Solution Approach 2:
The patent performs preliminary clustering on partial feature sets before final frame selection and labeling. This preliminary action organizes the data structure in advance, making subsequent frame-level analysis more efficient and improving overall productivity while maintaining measurement precision.
Data Source
AI summary
To make operation of generating training data more efficient in machine learning for events in data series. An information processing apparatus includes: a division section that divides, into a plurality of partial feature sets, a set of features constituting a feature series corresponding to each of a plurality of data series; a clustering section that clusters, into a plurality of clusters, a group of the plurality of partial feature sets which has been obtained from the plurality of data series; and a selection section that selects target data to be provided with a ground truth label, from a set of pieces of data which corresponds to at least one of the plurality of partial feature sets included in each of the plurality of clusters and which constitutes at least part of one of the plurality of data series.