Technology Category
Computer Vision: Representation Learning for VisionMachine Learning: Life-Long and Continual LearningIntelligent Robots: Multimodal Perception & Sensor Fusion
Multi-Level Feature Fusion for Continual Learning in Visual Quality Inspection
🤖 AI Summary
This work addresses the challenges of catastrophic forgetting and high computational cost in dynamic visual quality inspection scenarios—such as remanufacturing—where product types and defect patterns frequently change. The authors propose a multi-level feature fusion method built upon a pre-trained network, which enables efficient fine-tuning of representations at varying depths. By selectively updating only a small subset of parameters, the approach substantially reduces trainable parameters while effectively mitigating catastrophic forgetting. The method balances performance retention and generalization capability in continual learning settings, achieving accuracy comparable to end-to-end training across multiple inspection tasks, while significantly enhancing robustness to novel defects and products.
Technology Category
Application Category
📝 Abstract
Deep neural networks show great potential for automating various visual quality inspection tasks in manufacturing. However, their applicability is limited in more volatile scenarios, such as remanufacturing, where the inspected products and defect patterns often change. In such settings, deployed models require frequent adaptation to novel conditions, effectively posing a continual learning problem. To enable quick adaptation, the necessary training processes must be computationally efficient while still avoiding effects like catastrophic forgetting. This work presents a multi-level feature fusion (MLFF) approach that aims to improve both aspects simultaneously by utilizing representations from different depths of a pretrained network. We show that our approach is able to match the performance of end-to-end training for different quality inspection problems while using significantly less trainable parameters. Furthermore, it reduces catastrophic forgetting and improves generalization robustness to new product types or defects.
Problem
Research questions and friction points this paper is trying to address.
continual learning
visual quality inspection
catastrophic forgetting
remanufacturing
model adaptation
Innovation
Methods, ideas, or system contributions that make the work stand out.
Multi-Level Feature Fusion
Continual Learning
Catastrophic Forgetting
Visual Quality Inspection
Parameter Efficiency
🔎 Similar Papers
No similar papers found.
