Work

Fast-SCNN and YOLO11n-SGDE: A Hybrid Deep Learning Approach for Precise IC Flaw Identification

Year: 2026

Type: conference-paper

Abstract: This paper presents a lightweight two-stage inspection architecture for real-time, edge-deployable IC flaw identification. Fast-SCNN rapidly generates a clean ROI via semantic segmentation, and YOLO11n-SGDE then performs defect detection on the cropped ROI using an SGDE design that combines GhostConv, DWConv, and ECA for efficient feature extraction. Results show that the proposed pipeline reduces background-induced false positives and achieves strong detection performance (77.1% [email protected] and 77.1% F1-score) with only 2.08M parameters, outperforming YOLOv5n, YOLOv8n, and YOLO11n, thereby demonstrating its suitability for real-time industrial AOI on resource-constrained devices.

Source: IEEE International Conference on Consumer Electronics - Taiwan (ICCE-Taiwan)

DOI: 10.1109/icce-taiwan71481.2026.11652843

Authors: Hui-An Wu, Feng-Cheng Lin

Institution: Feng Chia University

Language: English

Topic: Industrial Vision Systems and Defect Detection

Subfield: Industrial and Manufacturing Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: hybrid

OpenAlex ID: W7204114146

FWCI:

Cites: 6

Cited by:

Related to: 10