2025/4/16 | 專題演講
114/4/17(四) 國立陽明交通大學資訊工程學系 黃敬群 教授
- 演講時間：114/4/17（四）15:30-17:00
- 演講地點：國立清華大學工程一館107演講廳
- 演講者： 黃敬群 教授 (優貝克科技股份有限公司)
- 講題： 「結合動態、保真與品質的盲人臉部修復方法」(Dynamic Blind Face Restoration: Bridging Fidelity and Quality for Real-World Applications)
- 大綱：
Face restoration in the wild is a challenging task, especially when the input comes from low-quality surveillance systems. This talk introduces a dynamic blind face restoration framework designed to handle open-set cross-resolution face recognition (CRFR) problems.
The method targets key challenges including:
- Resolution discrepancy between high-res galleries and low-res probes
- Large pose and texture variations in low-res faces
- Arbitrary image sizes and unknown degradations
We propose a unified solution that includes:
- Towards Understanding Cross Resolution Feature Matching for Surveillance Face Recognition
- Lipschitz-Conditioned for Resolution Robust Face Recognition
- Arbitrary-Resolution and Arbitrary-Scale Face Super-Resolution with Implicit Representation Networks
Theoretical analysis and extensive experiments on benchmark datasets such as SCface, CRLFW, and QMUL-Tinyface demonstrate the effectiveness of the proposed method.
Join us to explore how this framework bridges fidelity, visual quality, and real-world applicability in surveillance scenarios.