Super-Resolution & Upscaling¶
Image & 2D Art
Open Source¶
-
APISR Discovered by AI
2024
An image and video upscaler designed to restore and enhance low-quality, low-resolution anime images and video sources by analyzing the anime production workflow.
animecvprsuper-resolutionupscaling -
AuraSR Discovered by AI
2024
A GAN-based super-resolution model designed for upscaling generated images, based on a variation of the GigaGAN architecture for image-conditioned upscaling.
gangigagansuper-resolutionupscaling -
Clarity AI Upscaler Discovered by AI
2024
An open-source AI image upscaler and enhancer designed as a free alternative to Magnific AI, supporting high-resolution image generation and detail enhancement.
diffusionenhancermagnificupscaler -
DRCT: Saving Image Super-resolution away from Information Bottleneck Discovered by AI
2024
A state-of-the-art image super-resolution model based on dense relation connections and vision transformers, designed to mitigate the information bottleneck in deep networks.
ntiresuper-resolutiontransformer -
HiT-SR: Hierarchical Transformer for Efficient Image Super-Resolution Discovered by AI
2024
An efficient image super-resolution framework that converts transformer-based networks into hierarchical transformers. It replaces fixed small windows with expanding hierarchical windows to capture multi-scale features and establish long-range dependencies with linear complexity.
efficientsuper-resolutiontransformerupscaling -
MambaIR Discovered by AI
2024
MambaIR is a state-space model (SSM) baseline for image restoration and super-resolution, leveraging Mamba's linear complexity to capture long-range dependencies. It addresses the trade-off between global receptive fields and computational efficiency in traditional CNN and Transformer architectures.
mambastate-space-modelsuper-resolutionupscaling -
MambaIRv2 Discovered by AI
2024
An advanced image restoration and super-resolution model that equips the State Space Model (Mamba) with non-causal modeling capabilities. It overcomes the causal scanning limitations of vanilla Mamba to fully utilize spatial pixels and achieve state-of-the-art restoration quality.
mambarestorationstate-space-modelupscaling -
neosr Discovered by AI
2024
An open-source training framework for real-world single-image super-resolution models, optimized for fast training iterations, low GPU requirements, and practical restoration of photos and anime.
animesuper-resolutiontraining-frameworkupscaling -
OSEDiff: One-Step Effective Diffusion Network for Real-World Image Super-Resolution Discovered by AI
2024
An efficient one-step diffusion network for real-world image super-resolution that reduces inference steps and controls stochasticity to produce high-quality, consistent details.
diffusionneuripsone-stepsuper-resolution -
PiSA-SR: Pixel-level and Semantic-level Adjustable Super-resolution: A Dual-LoRA Approach Discovered by AI
2024
A diffusion-based super-resolution framework that uses a Dual-LoRA approach to separate and adjust pixel-level fidelity and semantic-level perceptual quality during inference.
cvprdiffusionlorasuper-resolution -
SeD Discovered by AI
2024
A semantic-aware discriminator framework for generative adversarial network (GAN) based image super-resolution. It incorporates semantic priors to guide the discriminator, mitigating virtual textures and improving the perceptual quality of reconstructed images.
gansemanticsuper-resolution -
StableSR: Exploiting Diffusion Prior for Real-World Image Super-Resolution Discovered by AI
2024
A generic image super-resolution method that leverages the generative priors of Stable Diffusion to achieve high-fidelity, arbitrary-scale upscaling.
diffusionijcvstable-diffusionsuper-resolution -
SUPIR Discovered by AI
2024
A groundbreaking image restoration and super-resolution method that leverages large-scale generative priors and model scaling to achieve highly realistic upscaling.
diffusionrestorationsdxlupscaling -
TSD-SR Discovered by AI
2024
A one-step diffusion framework for real-world image super-resolution that utilizes Target Score Distillation and a Distribution-Aware Sampling Module to achieve high-quality restoration with fast inference.
diffusiondistillationone-stepsuper-resolution -
CCSR Discovered by AI
2023
A diffusion-based super-resolution method designed to improve the content consistency and stability of upscaled images while accelerating the generation process.
consistencydiffusionsuper-resolution -
DAT Discovered by AI
2023
A transformer model for image super-resolution that aggregates features across spatial and channel dimensions to obtain powerful representation capabilities.
attentionsuper-resolutiontransformer -
DiffBIR Discovered by AI
2023
A general blind image restoration and super-resolution framework that leverages generative diffusion priors. It decouples the restoration process into degradation removal and detail regeneration stages to handle complex real-world image degradations.
diffusionrestorationsuper-resolution -
HAT Discovered by AI
2023
Hybrid Attention Transformer (HAT) combines channel attention and window-based self-attention to activate more input pixels for high-quality image restoration and super-resolution. It effectively exploits the potential of Transformer architectures in low-level vision tasks.
attentionsuper-resolutiontransformerupscaling -
OmniSR Discovered by AI
2023
A lightweight image super-resolution framework that uses omni-dimensional self-attention and dense interaction principles to achieve high-quality upscaling with low computational cost.
attentionlightweightsuper-resolution -
PASD Discovered by AI
2023
A pixel-aware Stable Diffusion framework designed for realistic image super-resolution and personalized stylization. It introduces a pixel-aware attention mechanism to maintain faithful pixel-wise image structures while leveraging strong generative priors.
diffusionstable-diffusionsuper-resolution -
ResShift Discovered by AI
2023
An efficient diffusion model for image super-resolution that uses a residual shifting mechanism to drastically reduce the number of sampling steps required during inference.
diffusionefficientsuper-resolution -
SeeSR Discovered by AI
2023
A semantics-aware real-world image super-resolution approach that utilizes a degradation-aware prompt extractor to preserve semantic fidelity in diffusion-based upscaling.
diffusionsemanticssuper-resolution -
SinSR Discovered by AI
2023
A diffusion-based image super-resolution method that achieves high-quality upscaling in a single step by distilling multi-step diffusion models.
diffusiondistillationone-stepsuper-resolution -
SRFormer Discovered by AI
2023
SRFormer introduces Permuted Self-Attention (PSA) to enable large-window self-attention in Transformer-based image super-resolution models with significantly reduced computational overhead. It achieves state-of-the-art performance in classical, lightweight, and real-world image super-resolution tasks.
attentionsuper-resolutiontransformerupscaling -
StableSR Discovered by AI
2023
A method that exploits the generative prior of pre-trained text-to-image diffusion models for real-world blind image super-resolution while preserving fidelity.
diffusionstable-diffusionsuper-resolution -
AnimeSR: Learning Real-World Super-Resolution Models for Animation Videos Discovered by AI
2022
A real-world super-resolution model specifically designed for animation videos, which learns degradation operators from real low-quality anime videos to synthesize realistic training data.
animesuper-resolutionvideo-sr -
HAT Discovered by AI
2022
A hybrid attention transformer for image super-resolution that combines channel attention and window-based self-attention to activate more input pixels for better reconstruction.
attentionsuper-resolutiontransformer -
Real-CUGAN Discovered by AI
2022
An anime-focused super-resolution model developed by Bilibili AI Lab, optimized for restoring and enhancing anime-style illustrations and video frames.
animesuper-resolutionupscaling -
Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration Discovered by AI
2022
An image restoration and super-resolution model based on the SwinV2 Transformer, specifically optimized for compressed images and video game textures.
compressioneccvsuper-resolutiontransformer -
2022
A free cross-platform desktop GUI for AI image upscaling (built on Real-ESRGAN and related models), a convenient front-end for batch-upscaling textures and illustrations.
desktopreal-esrganupscaling -
2021
A blind face-restoration model from Tencent ARC that recovers realistic facial detail from degraded images, often paired with upscalers to clean up portraits and character art.
face-restorationgan -
2021
A practical blind super-resolution model with a dedicated anime model variant, widely used to upscale textures and anime frames while preserving line sharpness.
animerestorationupscaling -
Restormer Discovered by AI
2021
An efficient transformer model designed for high-resolution image restoration, capturing both local and global features to handle tasks like super-resolution, deblurring, and denoising.
efficientrestorationsuper-resolutiontransformer -
SwinIR Discovered by AI
2021
A highly influential image restoration baseline model based on the Swin Transformer, achieving state-of-the-art performance in super-resolution, denoising, and compression artifact reduction.
restorationsuper-resolutionswintransformer -
Anime4K Discovered by AI
2019
A high-quality, real-time upscaling algorithm designed for anime video playback, optimizing performance and speed to preserve original content.
animereal-timeupscalingvideo -
2015
A CNN-based super-resolution and denoiser specialized for anime-style art, the long-standing baseline for upscaling illustrations and anime stills.
animeupscaling
Proprietary / Industry Tools¶
-
Krea AI Discovered by AI
2023
A creative AI suite featuring a powerful real-time upscaler and enhancer capable of upscaling images up to 22K resolution while restoring fine details and textures.
enhancerproprietaryreal-timeupscaler -
Krea AI Upscaler Discovered by AI
2023
Krea AI is a comprehensive creative AI suite featuring a powerful real-time upscaler and enhancer. It intelligently increases image resolution up to 22K while adding photorealistic details, making it highly suitable for concept art and production workflows.
proprietaryreal-timesuper-resolutionupscaling -
Magnific AI Discovered by AI
2023
A premium generative AI upscaler and enhancer that intelligently adds high-resolution details, textures, and sharpness to images and illustrations.
enhancergenerativeproprietaryupscaler -
Topaz Gigapixel AI Discovered by AI
2018
An industry-standard proprietary desktop application for image upscaling and super-resolution. It uses deep learning models trained on diverse datasets to add realistic details and sharpness while enlarging images up to 600%.
desktopproprietarysuper-resolutionupscaling
