Multi-Level Feature Fusion for Continual Learning in Visual Quality Inspection
This work addresses the need for computationally efficient and robust continual learning in domain-specific visual inspection tasks, representing an incremental improvement over existing methods.
The paper tackles the problem of adapting deep neural networks for visual quality inspection in volatile manufacturing scenarios like remanufacturing, where products and defects change frequently, by proposing a multi-level feature fusion approach that matches end-to-end training performance with significantly fewer trainable parameters, reduces catastrophic forgetting, and improves generalization robustness.
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.