Continual Error Correction on Low-Resource Devices
Kirill Paramonov, Mete Ozay, Aristeidis Mystakidis, Nikolaos Tsalikidis, Dimitrios Sotos, Anastasios Drosou, Dimitrios Tzovaras, Hyunjun Kim, Kiseok Chang, Sangdok Mo, Namwoong Kim, Woojong Yoo, Jijoong Moon, Umberto Michieli
TL;DR
This work tackles prediction errors in AI models deployed on resource-constrained devices by proposing a continual, few-shot error-correction framework. It combines server-side knowledge distillation from foundation models with an on-device prototype-based classifier that adapts through user-provided corrections without backpropagation. Prototypes per class are updated in a lightweight manner, enabling over 50% one-shot error correction on Food-101 and Flowers-102 while maintaining minimal forgetting ($<0.02 ext{%}$) and low computational overhead. The system is validated on image classification and object detection tasks, including an Android demonstration, demonstrating practical offline, private, and real-time error correction on consumer hardware. Overall, the approach offers a resource-efficient path to personalized, robust vision on edge devices with minimal user effort.
Abstract
The proliferation of AI models in everyday devices has highlighted a critical challenge: prediction errors that degrade user experience. While existing solutions focus on error detection, they rarely provide efficient correction mechanisms, especially for resource-constrained devices. We present a novel system enabling users to correct AI misclassifications through few-shot learning, requiring minimal computational resources and storage. Our approach combines server-side foundation model training with on-device prototype-based classification, enabling efficient error correction through prototype updates rather than model retraining. The system consists of two key components: (1) a server-side pipeline that leverages knowledge distillation to transfer robust feature representations from foundation models to device-compatible architectures, and (2) a device-side mechanism that enables ultra-efficient error correction through prototype adaptation. We demonstrate our system's effectiveness on both image classification and object detection tasks, achieving over 50% error correction in one-shot scenarios on Food-101 and Flowers-102 datasets while maintaining minimal forgetting (less than 0.02%) and negligible computational overhead. Our implementation, validated through an Android demonstration app, proves the system's practicality in real-world scenarios.