DiffBIR
Blind image restoration with generative diffusion priors: super-resolution, denoising, face restoration.
About
DiffBIR (Blind Image Restoration with Generative Diffusion Prior) splits restoration into two stages: a restoration module removes degradations, and an IRControlNet (a ControlNet variant trained on clean condition images) regenerates realistic detail on top of a frozen latent diffusion model. A region-adaptive guidance scale at inference lets users trade fidelity for realness without retraining. One pipeline covers blind super-resolution, blind denoising and blind face restoration.
The code is Apache 2.0 with pretrained models published (ECCV 2024; ~4,100 GitHub stars), and it runs anywhere PyTorch does. Like all generative restorers it can hallucinate plausible-but-wrong detail, so it suits enhancement and upscaling workflows rather than forensic or measurement uses.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- Provider-reported (paper, arXiv 2308.15070): superiority over SOTA approaches on blind super-resolution, face restoration and denoising across synthetic and real-world datasets
- No single headline benchmark number; evals are author-reported comparisons
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (TIER verified, metrics-v1). Evidence: ECCV 2024 paper with broad synthetic+real-world evals (arXiv 2308.15070); Apache 2.0 repo, ~4,100 GitHub stars; widely used as a restoration baseline in later work. Strengths: permissive license, unified multi-task pipeline, strong real-world track record. Docked for: (1) headline claims are author-reported with no independent benchmark; (2) generative hallucination risk inherent to the method. License: Apache 2.0.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · ECCV 2024 paper (arXiv 2308.15070) with synthetic + real-world evals
- · Apache 2.0 repo, ~4,100 GitHub stars (verified 2026-10-16)
- · Widely used as a restoration baseline in later work