Dynamic Feature Selection for Joint Probabilistic Image Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for joint recognition of semantic concepts in images are inefficient due to the computational expense of determining all desired features, especially when resources are limited, and fail to exploit interdependencies between concepts, leading to suboptimal classification performance.
Innovation Solution
A computer-implemented method that dynamically selects feature types for joint classification by estimating the improvement in classification accuracy and computation cost, using a probabilistic joint classifier to prioritize features based on direct and indirect accuracy improvements, and normalizing these improvements with computation costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all desired features are determined before starting the inferencing process, then classification accuracy is improved, but computational cost and processing time increase significantly
Solution Approach 1:
The patent implements dynamic feature selection where the set of features to be extracted is determined adaptively during the inferencing process rather than being fixed beforehand. The system dynamically selects features based on current image characteristics and classification needs, allowing the feature extraction process to respond to varying computational constraints and accuracy requirements in real-time
Solution Approach 2:
The patent segments the feature extraction process into multiple stages, extracting and evaluating features incrementally rather than processing all features simultaneously. This allows the system to divide the computational task into manageable portions, evaluating feature contributions progressively and stopping when sufficient accuracy is achieved, thereby reducing overall computational cost
2Measurement precision
If multiple feature types are used for joint recognition, then classification accuracy is improved, but computational resources are consumed excessively
Solution Approach 1:
The patent applies local quality by selecting and extracting only the specific feature types most relevant to each particular classification task and image content. Rather than uniformly processing all feature types across all concepts, the system identifies and extracts features locally tailored to the specific recognition objectives, reducing unnecessary computational overhead while maintaining accuracy for the most important classification dimensions
3Adaptability or versatility
If feature calculation is performed for all feature types, then comprehensive recognition is achieved, but recognition time is significantly delayed
Solution Approach 1:
The patent performs preliminary actions by pre-determining which features are most likely to be useful based on the classification task and image characteristics before extracting all features. This preliminary feature selection allows the system to prepare and prioritize feature extraction in advance, reducing the time required during actual recognition by avoiding unnecessary feature calculations
Data Source
AI summary
A method of jointly classifying a plurality of objects in an image using a feature type selected from a plurality of feature types determines classification information for each of the plurality of objects in the image by applying a predetermined joint classifier to at least one feature of a first type. The feature is generated from the image using a first feature extractor, the classification information being based on a probability of each of a plurality of possible classifications. The method estimates, for each of the feature types, an improvement in an accuracy of classification for each of the plurality of objects. The method selects features of a further type, from the plurality of feature types, according to the estimated improvement in the accuracy of the classification of each of the objects, and classifies the plurality of objects in the image using the selected features of the further type.