This paper provides a review of the NTIRE 2026 challenge on real-world face restoration, highlighting the proposed solutions and the resulting outcomes. The challenge focuses on generating natural and realistic outputs while maintaining identity consistency. Its goal is to advance state-of-the-art solutions for perceptual quality and realism, without imposing constraints on computational resources or training data. Performance is evaluated using a weighted image quality assessment (IQA) score and employs the AdaFace model as an identity checker. The competition attracted 96 registrants, with 10 teams submitting valid models; ultimately, 9 teams achieved valid scores in the final ranking. This collaborative effort advances the performance of real-world face restoration while offering an in-depth overview of the latest trends in the field.
Figures & tables
Figure 1: MiPlusCV adopts a two-stage pipeline that combines OSDFace-based coarse restoration with a Z-Image-based one-step detail enhancement stage.
Figure 2: Overview of the CEVI-KLETech pipeline. A semantic-aware wavelet correction block is inserted between the diffusion and naturalness stages.
Figure 3: Overview of the HONORAICamera pipeline.
Figure 4: YuFans combines a one-step SDFace restoration stage with CLIPIQA-guided pixel optimization at test time.
Figure 5: Overview of DeSC-Face. The degraded image is encoded into degraded latent tokens, which are used both as the main condition for the LoRA-adapted FLUX.2 backbone and as the input to the structured control branch. The scene-token stream is iteratively restored and then decoded into the final output image.
Figure 6: Architecture diagram of the DiffBIR v2.1 two-stage pipeline used by NTR. Stage 1 removes degradations with SwinIR, and Stage 2 synthesizes facial textures with IRControlNet conditioned on Stable Diffusion 2.1.
Figure 7: Overall architecture and training objective of MaDENN. The baseline CodeFormer architecture is extended with identity-preserving and ROI-aware supervision, while low-quality inputs are synthesized with a second-order Real-ESRGAN degradation pipeline.
Figure 9: The workflow of PRIDE-Face. GFPGAN provides the structural prior in the first stage, while DiffBIR synthesizes high-fidelity details in the second stage.
Figure 10: The BVI extends TADSR with a residual noise refiner and a detail-aware training strategy inside the one-step student branch.