PulseAugur
EN
LIVE 15:57:37

CFRNet achieves real-time face restoration on embedded NPUs

By PulseAugur Editorial · [2 sources] ·

Researchers have developed CFRNet, a new model for real-time blind face restoration on consumer embedded NPUs. This model utilizes a novel Cycle-Consistent Fixed-Point Training (CCFP) method, which trains the network to act as a fixed-point operator, improving image quality without adding inference cost. CFRNet achieves superior perceptual scores and PSNR/SSIM metrics compared to baselines retrained under similar deployment constraints. The model demonstrates efficient performance, running in approximately 23ms per cycle on a HiSilicon Hi3402 NPU and is capable of real-time operation on in-car driver-monitoring systems. AI

IMPACT Enables high-quality face restoration on low-power, embedded devices, potentially improving real-time applications like driver monitoring.

RANK_REASON The cluster contains an academic paper detailing a new model and training method for computer vision tasks.

Read on arXiv cs.CV →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

CFRNet achieves real-time face restoration on embedded NPUs

COVERAGE [2]

  1. arXiv cs.CV TIER_1 English(EN) · Fuchen Li, Xinyang Wang, Yahui Zhang, Yuhan Chen, Jiahong Guo, Zhuohan Qin, Wenbo Ma ·

    CFRNet: Cycle-Consistent Fixed-Point Training for Real-Time Blind Face Restoration on Consumer Embedded NPUs

    arXiv:2606.06850v1 Announce Type: new Abstract: Blind face restoration on consumer devices has to balance image quality against speed and memory. Strong methods such as GFPGAN and CodeFormer give good perceptual quality, but they rely on large pretrained generative priors and on …

  2. arXiv cs.CV TIER_1 English(EN) · Wenbo Ma ·

    CFRNet: Cycle-Consistent Fixed-Point Training for Real-Time Blind Face Restoration on Consumer Embedded NPUs

    Blind face restoration on consumer devices has to balance image quality against speed and memory. Strong methods such as GFPGAN and CodeFormer give good perceptual quality, but they rely on large pretrained generative priors and on operators such as attention, codebook lookup, an…

RELATED ENTITIES

RELATED TOPICS