Technology Category
Computer Vision: Adversarial Attacks & RobustnessMachine Learning: Calibration & Uncertainty QuantificationIntelligent Robots: Multimodal Perception & Sensor Fusion
A Dataset and Baseline for Deep Learning-Based Visual Quality Inspection in Remanufacturing
🤖 AI Summary
In remanufacturing, the high diversity of component types and defect patterns renders visual quality inspection heavily reliant on manual labor, while existing deep neural networks exhibit insufficient generalization to unseen components or defect types. Method: We introduce the first benchmark image dataset specifically designed for gearbox remanufacturing, enabling rigorous evaluation of cross-component and cross-defect distribution shift generalization. We further propose a contrastive regularization loss that explicitly enforces intra-class compactness and inter-class separability in the feature space, thereby enhancing zero-shot and few-shot recognition capability for novel components and defects. Contribution/Results: Experiments demonstrate that our method significantly outperforms baseline models on the new dataset, achieving an average classification accuracy improvement of 9.2%. This advancement substantially improves the robustness and transferability of visual inspection systems in remanufacturing contexts.
Technology Category
Application Category
📝 Abstract
Remanufacturing describes a process where worn products are restored to like-new condition and it offers vast ecological and economic potentials. A key step is the quality inspection of disassembled components, which is mostly done manually due to the high variety of parts and defect patterns. Deep neural networks show great potential to automate such visual inspection tasks but struggle to generalize to new product variants, components, or defect patterns. To tackle this challenge, we propose a novel image dataset depicting typical gearbox components in good and defective condition from two automotive transmissions. Depending on the train-test split of the data, different distribution shifts are generated to benchmark the generalization ability of a classification model. We evaluate different models using the dataset and propose a contrastive regularization loss to enhance model robustness. The results obtained demonstrate the ability of the loss to improve generalisation to unseen types of components.
Problem
Research questions and friction points this paper is trying to address.
Automating visual quality inspection for remanufactured components using deep learning
Addressing generalization challenges with new product variants and defect patterns
Improving model robustness for unseen component types through contrastive regularization
Innovation
Methods, ideas, or system contributions that make the work stand out.
Novel image dataset for gearbox component inspection
Contrastive regularization loss to enhance model robustness
Benchmark generalization with different train-test splits
🔎 Similar Papers
No similar papers found.
