Stepwise Object Recognizer Training for Dynamic Category Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition systems face challenges in efficiently classifying targets in images, particularly when new categories are introduced or when targets are not similar to learned ones, leading to inefficiencies in training and recognition processes.
Innovation Solution
A recognizer training apparatus that includes an obtainer and a controller, which trains a first object recognizer with a multilayer structure of stepwise determiners, allowing for stepwise classification from higher to lower layers and adding new categories, and constructs a second object recognizer when necessary, to recognize targets without relying on learned targets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a first object recognizer with multilayer stepwise determiners is used for classification, then recognition accuracy is improved, but training time increases when new categories need to be added
Solution Approach 1:
The object recognizer is divided into multiple independent determiners operating at different layers (first determiner, second determiner, etc.). Each determiner handles specific classification tasks independently, allowing the system to process complex recognition tasks in segmented steps rather than retraining the entire model when new categories are introduced.
Solution Approach 2:
The system dynamically selects which determiner to use based on the recognition needs. When a target cannot be classified by lower-layer determiners, higher-layer determiners are activated. This dynamic adaptation allows the system to handle new categories efficiently without fixed retraining schedules.
2Adaptability or versatility
If the first object recognizer is trained with new categories, then adaptability is improved, but device complexity increases
Solution Approach 1:
Multiple determiners are nested in a hierarchical structure where higher-layer determiners contain and build upon the functionality of lower-layer determiners. The second determiner operates within the framework established by the first determiner, creating a nested architecture that manages complexity through organized hierarchy.
Solution Approach 2:
The system adds a new dimension to classification by introducing higher-layer determiners that operate above the original classification level. When lower-layer determiners fail to classify targets, the system transitions to higher-dimensional classification space provided by upper-layer determiners.
3Adaptability or versatility
If a second object recognizer is constructed for unrecognized targets, then recognition coverage is improved, but productivity decreases
Solution Approach 1:
Instead of always using the full multilayer determination process, the system applies partial action by using only the necessary determiners for each specific case. Lower-layer determiners handle common cases quickly, while higher-layer determiners are activated only when needed, avoiding excessive processing for simple recognition tasks.
Solution Approach 2:
The higher-layer determiners act as intermediaries between the basic classification system and complex unrecognized targets. When lower-layer determiners cannot classify a target, the higher-layer determiners mediate by providing additional classification capabilities without requiring complete system reconfiguration.
Data Source
AI summary
An information processing apparatus includes a communicator and a controller. The communicator obtains an image. The controller trains a first object recognizer. The first object recognizer consists of a plurality of stepwise determiners in a multilayer structure. A top-layer determiner classifies a target in the image into one of categories. A lower-layer determiner classifies the target in a category determined by a stepwise determiner in a higher layer into a lower category. The first object recognizer recognizes the target by classifying the target stepwise from a higher layer to a lower layer. The controller causes the stepwise determiners to classify the target in the image obtained by the communicator from a higher layer to a lower layer. The controller trains the first object recognizer by adding a new lower category to a higher category corresponding to a lower-layer determiner that cannot classify the target into an existing lower category.