Feature Vector Generation for Identity Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic control systems, such as those in robots, face limitations in computing power when detecting and recognizing human faces in personalized home environments, particularly under challenging conditions like large pose variations and difficult lighting.
Innovation Solution
An object detecting device generates a feature vector based on a feature map rather than directly from a picture, reducing computing resource costs and improving speed, and transmits this vector to a server for identification, thereby optimizing data transfer and processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If feature vector is generated directly from picture, then identification accuracy can be maintained, but computing resource cost and processing time increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-generating feature maps from pictures using a neural network model before actual identification occurs. These feature maps are stored and reused for multiple identification tasks, avoiding repeated direct processing of original pictures. This preliminary processing step reduces the computational burden during real-time identification while maintaining accuracy.
Solution Approach 2:
The patent introduces feature maps as an intermediary between pictures and feature vectors. Instead of directly converting pictures to feature vectors (which is computationally expensive), the system first converts pictures to feature maps, then derives feature vectors from these intermediate representations. This intermediary step significantly reduces computing resource consumption while preserving identification accuracy.
2Measurement precision
If deep neural networks are used for face detection and recognition, then accuracy under challenging conditions improves, but limitations in computing power become more pronounced
Solution Approach 1:
The patent extracts the essential features from pictures by generating feature maps that capture key characteristics needed for identification. Instead of processing entire high-resolution pictures through complex neural networks repeatedly, the system extracts and stores feature representations that contain the necessary identification information, significantly reducing computational requirements while maintaining recognition accuracy.
Solution Approach 2:
The patent changes the parameter representation from raw picture data to compressed feature map representations. By transforming the data from high-dimensional pixel values to lower-dimensional feature vectors derived from feature maps, the system reduces computational complexity and memory requirements while preserving the essential information needed for accurate face recognition under challenging conditions.
Data Source
AI summary
In an embodiment, the disclosure includes an object detecting device. The object detecting device is configured to execute the instructions to: obtain a first picture comprising a first object at a first time instant; determine a first feature pattern of the first object based on the first picture; generate a first feature map of the first object based on the first feature pattern; generate a first feature vector of the first object based on the first feature map; and send the first feature vector to a server. In this embodiment, the first feature vector is generated based on the first feature map by the object detecting device rather than starting another process of directly generating the first feature map based on the first picture. Therefore, the speed and the computing resource cost of generating the feature vector may be better.