Multiclass Object Classifier Using Binary Pre-filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing monitoring systems face challenges in accurately classifying residents due to the inclusion of residents in both positive and negative sample sets during training, leading to incorrect predictions and reduced accuracy.
Innovation Solution
The system initially trains binary classifiers for each resident using only resident images, then generates a multiclass classifier by excluding images of residents and similar individuals from the negative sample set during training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a multiclass classifier is trained using both resident images (positive samples) and unknown person images (negative samples), then the classifier can recognize both residents and unknown persons, but residents may be incorrectly included in the negative sample set leading to classification errors
Solution Approach 1:
The system performs preliminary classification using binary classifiers before training the multiclass classifier. Binary classifiers are trained on resident images first, then used to identify and exclude resident images from the unknown person set before multiclass classifier training, preventing classification errors
Solution Approach 2:
Binary classifiers serve as an intermediary step between raw image data and the final multiclass classifier. They act as a filtering mechanism to identify and remove resident images from the negative sample set, ensuring clean training data for the multiclass classifier
2Productivity
If a classifier is trained with images of a resident as both positive and negative samples, then training data is maximally utilized, but the classifier becomes unable to accurately classify the resident
Solution Approach 1:
Binary classifiers are trained in advance on resident images before multiclass classifier training. This preliminary action ensures residents are correctly identified and excluded from negative samples, maintaining classification reliability while efficiently utilizing training data
3Adaptability or versatility
If the unknown person set includes a person that looks very similar to the resident, then the training data is more diverse and representative, but the classifier becomes unable to accurately distinguish between the similar persons
Solution Approach 1:
Binary classifiers are trained on resident images before multiclass classifier training. These binary classifiers serve as a preliminary filter to identify and exclude images of persons similar to residents from the unknown person set, preventing false negatives while maintaining training data diversity
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for objection classification. One of the methods includes: obtaining a first set of images of objects that have a likelihood of being at a property that satisfies a likelihood threshold; generating, for each object, a binary classifier from a set of images of the respective object; determining, using at least one of the binary classifiers, that an image of an unknown object was classified as an object from the objects; in response to determining, using the binary classifiers, that the image of the unknown object was classified as an object from the objects, selecting a second set of images of unknown objects that does not include the image; and generating a multiclass classifier for use classifying objects using i) the first set as respective classes and ii) the second set that does not include the image.