Wafer flaw detection method based on deep learning

The invention belongs to the technical field of wafer detection, and provides a wafer flaw detection method based on deep learning. A teacher network and a student network are obtained through improvement based on an existing YOLOv7 target detection algorithm, and the teacher network and the student...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: LAN HAO, WANG RUI, XIA XINDONG, YANG BIAO, QI LULU, YANG CHANGCHUN, JIANG FENG
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention belongs to the technical field of wafer detection, and provides a wafer flaw detection method based on deep learning. A teacher network and a student network are obtained through improvement based on an existing YOLOv7 target detection algorithm, and the teacher network and the student network form a double-channel teacher-student network. The input image is trained through a teacher network to obtain a ubiquitous flaw area; ubiquitous flaw information obtained by the teacher network in the training process is migrated to the student network through distillation knowledge; performing student network training on the input image to obtain subdivided flaws in the ubiquitous flaw area; in the final test stage, only the trained student network is reserved, and whether the wafer is normal or not and the defect type of the wafer are obtained by inputting images. According to the method, the real-time performance of flaw subdivision can be ensured, the learning effect is ensured, and the subdivision fla