Object Detection Using Sentence Label Embeddings and Attention Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection methods are limited in utilizing sentence form labels, as they only consider word form labels, which restricts the accuracy and efficiency of object detection.
Innovation Solution
A machine learning apparatus and method that incorporates sentence label embeddings into the feature extraction process by partitioning object regions, creating attention matrices, and training object detection models using these embeddings to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sentence label embeddings are incorporated into the feature extraction process, then object detection accuracy is improved, but device complexity increases
Solution Approach 1:
The object region is partitioned into multiple sub-regions, and sentence labels are segmented and embedded into corresponding sub-region features. This segmentation allows the complex sentence label information to be distributed and integrated systematically across different parts of the object region, improving detection accuracy without overwhelming the model structure.
Solution Approach 2:
Sentence label embeddings are nested within the feature extraction process by integrating them into the sub-region features. The sentence label information is embedded as additional feature dimensions within the existing feature space, allowing multi-level information (visual features + linguistic features) to be combined in a hierarchical manner that improves accuracy while maintaining structural efficiency.
2Measurement precision
If attention matrices are created from sub-regions along bounding surfaces, then outer shape information is incorporated improving detection accuracy, but loss of time increases
Solution Approach 1:
Sub-regions along the bounding surfaces are extracted from the complete object region to create attention matrices. By focusing only on the boundary sub-regions rather than processing the entire object region, the method efficiently captures outer shape information while reducing the computational burden and processing time compared to analyzing all regions.
Solution Approach 2:
The attention mechanism is applied locally to sub-regions along the bounding surfaces rather than uniformly across the entire object region. This local application of attention allows the model to focus computational resources on the most informative boundary areas that contain critical outer shape information, improving detection accuracy while minimizing time loss.
Data Source
AI summary
A machine learning apparatus according to an embodiment includes a feature extractor configured to extract features from an object region of an image, a label processor configured to create sentence label embeddings from a sentence label corresponding to the object region, a first training data creator to extract first sub-features from a plurality of first sub-regions created by partitioning the object region, add the sentence label embeddings to the extracted first sub-features, and add the first sub-features added with the sentence label embeddings to the features of the object region, a second training data creator to extract a plurality of second sub-regions along a bounding surface of the object region, create an attention matrix from the second sub-regions, and create a training data by applying the attention matrix to the features of the object region, and a trainer to train an object detection model using the training data.