Parallel Deep Neural Network for Dynamic Article Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In logistics and production fields, existing deep learning systems require a long time to relearn when new articles are added, making it difficult to evolve the system without impairing actual operation.
Innovation Solution
An information processing device with a parallel deep neural network that performs inferences on multiple articles simultaneously, a new article determination unit to identify unlearned articles, and a new article learning unit to learn and add new deep neural network models without disrupting ongoing operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a multi-class deep neural network covering all varieties of articles is learned in advance, then the system can perform classification and position information detection on multiple types of articles, but the learning time becomes extremely long and the system cannot adapt to new articles without relearning all articles
Solution Approach 1:
The patent divides the single multi-class deep neural network into multiple independent single-class deep neural networks, each dedicated to recognizing one specific type of article. This segmentation allows each model to be trained separately and independently, significantly reducing the learning time required for each individual model while maintaining the ability to recognize multiple article types through parallel inference.
Solution Approach 2:
The patent implements a dynamic system where new single-class deep neural networks can be added to the parallel group as new article types are introduced. This dynamic architecture allows the system to adapt to new articles without relearning existing models, enabling continuous evolution of the recognition system while maintaining operational efficiency.
2Adaptability or versatility
If the deep neural network is relearned to cover new articles, then the system can recognize the new article type, but the actual operation is impaired and the learning process takes a long time
Solution Approach 1:
By segmenting the recognition system into independent single-class models, the patent enables individual model updates without affecting the entire system. When a new article type needs to be recognized, only the new single-class model for that specific article type needs to be trained and added, while existing models continue to operate unchanged, ensuring operational continuity.
Solution Approach 2:
The patent prepares new single-class deep neural network models in advance for new article types before they need to be recognized. These pre-trained models can be quickly integrated into the parallel system without disrupting ongoing operations, allowing the system to adapt to new articles with minimal operational impact.
3Productivity
If a parallel deep neural network with multiple single-class models is used, then the inference speed improves and new articles can be added easily, but the device complexity increases
Solution Approach 1:
The patent segments the recognition task into multiple independent single-class models that operate in parallel. While this increases the number of models, each model is simpler and more specialized, making them easier to train, manage, and update independently. The parallel architecture improves inference speed by simultaneously evaluating multiple article types without the computational burden of a single large multi-class model.
Solution Approach 2:
The patent creates a universal framework where each single-class deep neural network model serves a specific article type but follows the same architectural pattern and training procedure. This universality allows for standardized model deployment, management, and integration, reducing the operational complexity despite having multiple models. The system can handle any number of article types by simply adding or removing individual models as needed.
Data Source
AI summary
An information processing device includes a parallel deep neural network configured to input a captured image of an article to deep neural network models respectively corresponding to a plurality of articles and perform inferences about the plurality of articles in parallel using the deep neural network models, a new article determination unit configured to determine whether an article included in the image is an unlearned article based on learned model information about the articles and the image, and a new article learning unit configured to learn a deep neural network model corresponding to the article determined to be unlearned based on the image and initial model configuration information about the deep neural network model when the article included in the image is determined to be an unlearned article. The new article learning unit adds the learned deep neural network model to the deep neural network models.