In this work, we compare the performance of convolutional neural networks and support\nvector machines for classifying image stacks of specular silicon wafer back surfaces. In these image\nstacks, we can identify structures typically originating from replicas of chip structures or from\ngrinding artifacts such as comets or grinding grooves. However, defects like star cracks are also visible\nin those images. To classify these image stacks, we test and compare three different approaches. In the\nfirst approach, we train a convolutional neural net performing feature extraction and classification.\nIn the second approach, we manually extract features of the images and use these features to train\nsupport vector machines. In the third approach, we skip the classification layers of the convolutional\nneural networks and use features extracted from different network layers to train support vector\nmachines. Comparing these three approaches shows that all yield an accuracy value above 90%.\nWith a quadratic support vector machine trained on features extracted from a convolutional network\nlayer we achieve the best compromise between precision and recall rate of the class star crack with\n99.3% and 98.6%, respectively.
Loading....