Local Individual Recognition With Loop-Based Self-Learning for Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional manual and cloud-based systems for attendance monitoring, access control, and people counting are labor-intensive, time-consuming, prone to circumvention, and raise concerns about cost, privacy, and data security.
Innovation Solution
A local recogniser is trained with local data, updates and refines its training data using additional data from a remote recogniser, and performs self-learning through an iterative loop to recognize individuals, with periodic updates based on scheduling data, using CNNs for embedding and object recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If cloud-based biometric or facial recognition systems are used for recognising individuals, then recognition capability is provided, but concerns about cost, privacy, and data security arise
Solution Approach 1:
The system segments the recognition functionality into a local recogniser that processes images on-device, eliminating the need to transmit biometric data to cloud servers. This segmentation keeps sensitive data local, addressing security and privacy concerns while maintaining recognition capability through the local recogniser trained with local training data.
Solution Approach 2:
The local recogniser performs self-learning by automatically updating its training data with additional training data retrieved from a remote recogniser when individuals cannot be locally recognised. This self-service mechanism improves recognition accuracy over time without requiring manual intervention or cloud-based processing of sensitive data.
2Device complexity
If manual attendance tracking is used, then simplicity is maintained, but labour-intensive and time-consuming operations occur
Solution Approach 1:
The system replaces manual mechanical attendance tracking with an automated image-based recognition system. The local recogniser automatically captures images, identifies individuals, and records attendance, eliminating the need for manual tracking while maintaining system simplicity through local processing and minimal infrastructure requirements.
3Reliability
If cloud-based systems are deployed, then recognition functionality is provided, but labour-intensive operations and time-consuming processes occur
Solution Approach 1:
By segmenting the recognition process to occur locally on the device rather than in the cloud, the system eliminates network transmission delays and cloud processing bottlenecks. The local recogniser processes images immediately, significantly reducing processing time while maintaining recognition functionality.
4Measurement precision
If local recogniser with self-learning is used, then recognition accuracy is improved, but additional data retrieval and processing steps are required
Solution Approach 1:
The local recogniser implements self-learning by automatically retrieving additional training data when recognition fails and retraining itself. This self-service approach improves recognition accuracy without requiring manual intervention or complex external systems, as the recogniser autonomously manages its own training and improvement.
Solution Approach 2:
The system incorporates feedback loops where the local recogniser evaluates its own performance, identifies individuals that cannot be recognised, retrieves additional training data for those cases, and retrains. This feedback mechanism systematically improves recognition accuracy while managing processing steps through automated workflows.
Data Source
AI summary
A method for recognising individuals at a location may include: locally capturing images of individuals at the location; locally recognising individuals in the images by a local recogniser trained with local; training data for individuals previously recognised, or expected to be, at the location; for individuals that initially cannot be locally recognised, retrieving additional training data from a remote recogniser using query data extracted from the images by the local recogniser; updating the local training data with the additional training data; retraining the local recogniser with the updated local training data to locally recognise the individuals that initially could not be locally recognised, and wherein the local training data, query data and additional training data comprise embeddings extracted from images of individuals.