Surface Quality Automatic Inspection for Pharmaceutical Capsules Using Deep Learning

Capsules are commonly used as containers for most pharmaceutical products. Thus, the quality of a capsule is closely related to therapeutic effect products and patient health. At present, surface testing an essential task in actual production capsules. In this study, deep learning-based defect detection model, called CapsuleDet, proposed classify localize defects image sensor data from practical application. A guided filter-based enhancement method hybrid augmentation improving quantity raw data, respectively, mitigate low contrast issue enhance robustness model training. Deformable convolution module attentional fusion feature pyramid also improve by effectively utilizing semantic geometric information extracted maps catering with different shapes scales. The evaluation results on dataset demonstrate that achieves 92.91% mean average precision 22.16 frames per second. Moreover, its overall performance terms training time, size, accuracy, speed better than currently popular detectors.