Lightweight Industrial Image Classifier Based on Federated Few-Shot Learning

Abstract

Image classification using Convolutional Neural Networks (CNNs) is critical for broader industrial applications like defect detection. To protect sensitive data during industrial process, increasing institutions are highly interested in training CNN classifiers collaboratively with Federated Learning (FL). However, existing FL solutions cannot address the sample deficiency and heterogeneous learning resource issues at different practical institutions. In this paper, we present FLRN, a lightweight industrial image classifier based on our Federated Few-Shot Learning (FFSL) architecture. Results of extensive experiments considering different real-world FFSL scenarios indicate that, unlike the state-of-the-art few-shot learning method Relation Network (RN), FLRN performs well on not only FL participants with mutually isolated classes of samples but also external institutions with limited samples from unseen classes. Compared to RN with the predominating FedAvg-based FL deployment, FLRN manages to achieve as low as 29.6× less client-cloud communication, 5.2× less computation, and 22.0× less storage costs of clients.