Image Classification of PCBs and its Web Application (Flask)
Spotting defective printed circuit boards by eye is slow and error-prone. This article trains a lightweight image classifier to label PCBs as good or defective, then wraps it in a simple web app.
- A MobileNet model pre-trained on ImageNet, fine-tuned with transfer learning on 1,095 PCB images
- Real-time augmentation, early stopping and learning-rate scheduling to avoid overfitting
- A Flask web app with an upload page and a results page, built on Bootstrap templates
