 |
International Society of Science and Applied Technologies |
| 3D Convolutional Network-Based Monitor Flicker Defect Detection: A Structured Approach for Real-World Application | ||||
| Author | Jia-Hong Chou
|
|||
| Co-Author(s) | Fu-Kwun Wang; Alfian Nur Hidayat; Chin-Chung Wu; Tao-Peng Chu
|
|||
| Abstract | The reliability and quality control of consumer electronics are crucial in both the production process and outgoing quality control. Laptops, widely used for gaming and video streaming, are particularly affected by monitor issues such as flashing, lines, points, and flickering, which significantly impact user experience. This research proposes a novel framework integrating You Only Look Once version 8 (YOLOv8) to identify the monitor's location from image frames captured by an extended webcam and send the focused region to a well-trained Convolutional 3D (C3D) Network for flicker detection. Frames per second (FPS) and model accuracy are key factors in determining real-world applicability. Through experimental analysis, the C3D model is selected for its high FPS and superior performance compared to other existing 3D convolutional-based models. The proposed framework achieves high accuracy in flicker detection and has been successfully tested in real-time within an industrial environment at a well-known company.
|
|||
| Keywords | Flicker detection, real-time detection, temporal CNN-based model, product reliability, frames per second | |||
| Article #: RQD2025-257 | ||||
Proceedings of 30th ISSAT International Conference on Reliability & Quality in Design |