Developing a Resource-Constraint EdgeAI model for Surface Defect Detection

TL;DR

Edge devices face latency, privacy, and resource constraints that impede traditional cloud-trained models. The authors propose a lightweight, Xception-inspired architecture augmented with SqueezeNet-style reductions to enable on-device training and inference for PCB defect detection, and compare it against MobileNetV2, EfficientNetV2B0, and MobileViT-XXS. Without pretraining the proposed model achieves 73.45% test accuracy, demonstrating strong performance while highlighting that architectural design can beat parameter-count alone in resource-constrained settings; transfer learning further boosts results for some baselines. The findings suggest that on-device training is feasible for defect detection and can generalize to other edge-only ML tasks, guiding future work toward pretraining at scale and additional model-squeezing to broaden edge deployment.

Abstract

Resource constraints have restricted several EdgeAI applications to machine learning inference approaches, where models are trained on the cloud and deployed to the edge device. This poses challenges such as bandwidth, latency, and privacy associated with storing data off-site for model building. Training on the edge device can overcome these challenges by eliminating the need to transfer data to another device for storage and model development. On-device training also provides robustness to data variations as models can be retrained on newly acquired data to improve performance. We, therefore, propose a lightweight EdgeAI architecture modified from Xception, for on-device training in a resource-constraint edge environment. We evaluate our model on a PCB defect detection task and compare its performance against existing lightweight models - MobileNetV2, EfficientNetV2B0, and MobileViT-XXS. The results of our experiment show that our model has a remarkable performance with a test accuracy of 73.45% without pre-training. This is comparable to the test accuracy of non-pre-trained MobileViT-XXS (75.40%) and much better than other non-pre-trained models (MobileNetV2 - 50.05%, EfficientNetV2B0 - 54.30%). The test accuracy of our model without pre-training is comparable to pre-trained MobileNetV2 model - 75.45% and better than pre-trained EfficientNetV2B0 model - 58.10%. In terms of memory efficiency, our model performs better than EfficientNetV2B0 and MobileViT-XXS. We find that the resource efficiency of machine learning models does not solely depend on the number of parameters but also depends on architectural considerations. Our method can be applied to other resource-constraint applications while maintaining significant performance.

Figures (6)

• Figure 1: The proposed method to build a lightweight EdgeAI model.
• Figure 2: Comparison of Training Accuracy and Training Loss
• Figure 3: Resource Consumption Pattern
• Figure A.1: Comparison of Pre-trained models vs Our Non pre-trained model
• Figure A.2: Benchmark result with Xception (pre-trained on ImageNet) implemented on desktop