Dynamic Learning Model Selection for Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object recognition devices for autonomous driving struggle to accurately detect moving objects while maintaining low calculation loads, leading to decreased recognition accuracy and increased processing speed.
Innovation Solution
An object recognition system that determines the vehicle's scene using sensor information and selects a suitable learning model from multiple models based on the scene and external factors, such as brightness and weather conditions, to improve object recognition accuracy during driving.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If uniform object recognition is applied to all objects in driving scenes, then recognition accuracy of each object type can be improved, but processing speed decreases due to increased calculation load
Solution Approach 1:
The patent segments the object recognition task by dividing it into multiple learning models, each specialized for specific object types (e.g., pedestrian model, vehicle model, bicycle model). The system selects and applies only the relevant learning model based on the detected object category, rather than using a single uniform recognition process for all objects. This segmentation reduces unnecessary calculations while maintaining high recognition accuracy for each object type.
Solution Approach 2:
The patent implements dynamic selection of learning models based on the driving scene and object characteristics. The system dynamically determines which learning model to apply by analyzing scene information and object features, switching between different recognition approaches as needed. This dynamic adaptation allows the system to optimize processing speed by selecting simpler models when appropriate while maintaining accuracy when complex recognition is needed.
2Device complexity
If a single learning model is used for all object recognition tasks, then device complexity is reduced, but recognition accuracy decreases for specific object types
Solution Approach 1:
The patent creates a universal object recognition system that handles multiple object types through a unified framework. Instead of requiring separate dedicated systems for each object type, the system uses a single multi-functional architecture that can adaptively select and apply appropriate learning models for different objects (pedestrians, vehicles, bicycles, etc.), achieving both simplicity and high accuracy.
Solution Approach 2:
The patent changes the parameters of the recognition system by maintaining multiple learning models with different parameters and characteristics, then dynamically selecting the appropriate model based on scene parameters and object features. This allows the system to optimize recognition accuracy for each object type without permanently increasing system complexity, as only one model is actively used at a time.
3Measurement precision
If multiple learning models are maintained for different object types, then recognition accuracy for specific objects improves, but device complexity and memory requirements increase
Solution Approach 1:
The patent performs preliminary classification of objects in the driving scene before applying specific learning models. By first identifying the general category of object detected and then selecting the appropriate specialized model, the system avoids the complexity of managing and switching between multiple models continuously. The preliminary classification step simplifies model selection and reduces the computational overhead of model management.
4Productivity
If all objects are recognized uniformly without scene consideration, then processing speed is maintained, but recognition accuracy decreases in complex driving scenarios
Solution Approach 1:
The patent applies local quality by selecting different learning models based on the specific characteristics of the driving scene and object location. Instead of using a uniform recognition approach throughout, the system adapts the recognition method to local conditions - choosing appropriate models based on scene context, object position, and environmental factors. This localized adaptation maintains processing speed by avoiding unnecessary complex processing in simple scenarios while improving accuracy when needed.
Data Source
AI summary
An object recognition device 80 includes a scene determination unit 81, a learning-model selection unit 82, and an object recognition unit 83. The scene determination unit 81 determines, based on information obtained during driving of a vehicle, a scene of the vehicle. The learning-model selection unit 82 selects, in accordance with the determined scene, a learning model to be used for object recognition from two or more learning models. The object recognition unit 83 recognizes, using the selected learning model, an object in an image to be photographed during driving of the vehicle.