Tagged articles
super-resolution
40 articles · Page 1 of 1
PolaFormer++ introduces a polarity-aware linear attention mechanism with channel-wise spiky feature mapping (PaCS) that solves negative value loss and high entropy issues, achieving state-of-the-art results across six vision tasks including classification, detection, segmentation, diffusion, 3D synthesis, and super-resolution.
3D ReconstructionADE20KCOCO
0 likes · 16 min read


SuanNiJun 6, 2026 · Artificial Intelligence
JoyAI‑Echo introduces a cross‑modal audio‑visual memory bank, a three‑stage post‑training pipeline, and a Director Agent to enable consistent, high‑quality, real‑time generation of minute‑level videos, achieving up to 7.5× inference speedup and state‑of‑the‑art benchmark scores.
Director AgentJoyAI-Echoaudio-visual AI
0 likes · 13 min read

JoyAI‑Echo, an open‑source framework from JD, enables fully consistent five‑minute AI video generation with a single pass, offering non‑linear editing, high‑resolution output up to 1472×2560, and a suite of memory‑driven techniques that overcome the long‑video bottlenecks of earlier models.
AI Video GenerationDirector AgentJoyAI-Echo
0 likes · 12 min read

Xiaomi secured three top spots at CVPR 2026 NTIRE—first in Efficient Super‑Resolution with SPANV2, first in Portrait Restoration using a dual‑stage cascade, and second in Reflection Removal via RDNet‑XL and diffusion‑model distillation—showcasing hardware‑software co‑design, ultra‑fast inference, and novel algorithmic innovations.
diffusion model distillationefficient inferencehardware-software co-design
0 likes · 14 min read


AIWalkerApr 6, 2026 · Artificial Intelligence
The TPAMI 2025 paper introduces BIPNet, a unified burst‑image framework that tackles alignment, fusion, and upsampling challenges with edge‑enhanced alignment, pseudo‑burst feature fusion, and adaptive group upsampling, achieving state‑of‑the‑art results across super‑resolution, low‑light enhancement, and denoising while offering lightweight mobile variants.
BIPNetBurst Image ProcessingComputer Vision
0 likes · 13 min read

This article presents Ada-RefSR, a novel single‑step diffusion‑based reference super‑resolution framework that introduces a "Trust but Verify" paradigm, adaptive implicit correlation gating, and lightweight architecture to robustly suppress hallucinations and achieve state‑of‑the‑art performance on multiple benchmarks, while being suitable for mobile deployment.
Ada-RefSRICLR2026diffusion
0 likes · 10 min read

Kuaishou has had eight cutting‑edge papers accepted at the International Conference on Machine Learning 2025, covering breakthroughs in multimodal emotion modeling, monotonic probability learning, causal effect generalization, cascade ranking, multimodal LLM alignment, ultra‑low‑rate image compression, and visual autoregressive super‑resolution, with links to each work and accompanying code repositories.
AICausal InferenceRanking
0 likes · 13 min read

The article presents MaIR, a locality‑ and continuity‑preserving Mamba‑based model for image restoration, detailing its three‑stage architecture, novel scanning strategy, loss functions, experimental results on super‑resolution and denoising, and ablation studies, with links to the arXiv paper and source code.
Computer VisionDenoisingMamba
0 likes · 5 min read

The article presents MaIR, a Mamba‑based image restoration model that preserves locality and continuity, detailing its architecture, scanning strategies, loss functions, experimental results on super‑resolution and denoising, and an ablation study, while providing links to the arXiv paper and GitHub source code.
Computer VisionDenoisingMamba
0 likes · 5 min read


AIWalkerMar 25, 2025 · Artificial Intelligence
ContinuousSR introduces a Pixel-to-Gaussian paradigm that models images as continuous Gaussian fields, enabling arbitrary‑scale super‑resolution with 0.9 dB PSNR gains and up to 19.5× faster rendering compared to existing methods.
Arbitrary-Scale SRComputer VisionContinuousSR
0 likes · 5 min read

Introducing MambaIRv2, an image restoration backbone that replaces Mamba’s causal scanning with an attentive state‑space module, achieving single‑direction scanning, reduced parameters and computation, and superior performance on lightweight and classic super‑resolution, JPEG artifact removal, and denoising tasks, as validated by CVPR‑2025 results.
AttentionComputer VisionMambaIRv2
0 likes · 8 min read


AIWalkerFeb 6, 2025 · Artificial Intelligence
FluxSR introduces a novel single‑step diffusion approach for real‑world image super‑resolution built on the 12‑billion‑parameter FLUX.1‑dev model, employing Flow‑Trajectory Distillation, TV‑LPIPS and attention‑diversity losses to achieve high fidelity, reduced artifacts, and lower memory and compute costs.
Flow Distillationdiffusionimage restoration
0 likes · 16 min read

Xiaohongshu’s audio‑video architecture team secured second place in the RAIM challenge and first in the S‑UGC VQA challenge at CVPR NTIRE 2024 by improving generative image restoration with SUPIR, DeSRA and a Fusion model, and enhancing video quality assessment using LIQE, Q‑Align and FAST‑VQA, then deploying these methods for live‑stream denoising, intelligent transcoding and cloud‑based super‑resolution, achieving high PLCC/SROCC scores and up to 33 % bandwidth savings.
AICVPR NTIRE 2024Denoising
0 likes · 25 min read

Bilibili’s self‑supervised video super‑resolution system upgrades low‑resolution streams to 4K by using three parallel degradation‑branch networks—texture‑enhancing, line‑recovering, and noise‑removing—tailored to anime, game, and real‑world content, delivering sharper edges, finer textures, and measurable quality gains across its online playback pipeline.
AIBilibilideep learning
0 likes · 16 min read

The article details Tencent Music's Tianqin Lab research on audio quality AIGC, covering background upgrades, music separation techniques, real‑time super‑resolution, the industry‑first premium master‑track technology, and a Q&A on vocal separation, illustrating practical AI-driven audio enhancements.
AIGCMasteringMusic Separation
0 likes · 7 min read

This article reviews the latest research from Alibaba DAMO Academy on real-world image quality problems, covering spatial, temporal, and color enhancement methods, advanced quality assessment metrics, multimodal diffusion models, and future directions toward large‑model integration and lightweight deployment.
MOS regressiondeep learningimage enhancement
0 likes · 24 min read

At LiveVideoStackCon 2023, XiaoHongShu unveiled an edge‑cloud super‑resolution framework guided by the human‑perception‑aligned RedVQA model, which jointly optimizes video quality and bandwidth by generating a dedicated SR bitrate tier in the cloud and applying a lightweight SR algorithm on the client, achieving notable QoE gains and narrow‑band HD delivery.
AIbandwidth optimizationedge-cloud
0 likes · 25 min read

The LiveVideoStackCon 2023 session revealed how a human‑perception‑driven end‑cloud super‑resolution framework, AI‑based no‑reference video quality assessment, and rigorous AB‑testing methods can dramatically reduce video bandwidth while enhancing visual quality, illustrating the broader challenges and opportunities in modern audio‑video systems.
AB TestingAI assessmentaudio-video industry
0 likes · 13 min read

Bilibili’s Cloud Multimedia team uses AI‑driven pipelines to cut bandwidth costs while delivering low‑bitrate, high‑quality video, employing a QoE‑based decision engine, real‑time 4K super‑resolution for game streams, low‑rank reconstruction for narrow‑band HD, data‑driven HDR LUTs, and explores diffusion‑based restoration for legacy content.
AIBilibililow bitrate
0 likes · 27 min read

Youku’s Digital Media Lab built an AI‑powered restoration pipeline that uses multi‑frame super‑resolution, denoising, deblocking, deblurring and GAN‑based detail generation to automatically revive classic animated films, removing noise and artifacts while preserving fine lines, enabling high‑definition viewing for modern audiences.
AIDenoisinganimation
0 likes · 8 min read

The TaoMC2 framework, a two‑stage pipeline that augments BasicVSR++ with peak‑quality frames and deep residual blocks in Stage I and refines results with a SwinIR transformer in Stage II, leverages progressive training and transfer learning to boost PSNR to 33.16 dB and secured two championship titles and a runner‑up in the NTIRE 2022 video enhancement challenge.
NTIRE2022compression artifact removalsuper-resolution
0 likes · 26 min read

The article explains how BIGO RTC achieves high‑quality, low‑cost real‑time audio‑video communication by optimizing core video operators such as HEVC encoding, AI‑driven super‑resolution and HDR, and by employing scene‑adaptive techniques like device performance tuning, content‑adaptive encoding and AI‑based pre‑processing to meet diverse latency constraints.
AI AdaptationHDRHigh Quality
0 likes · 9 min read

This article explores the rapid growth of ultra‑HD video, the challenges it creates for cloud transcoding, and how Huawei Cloud leverages AI techniques such as super‑resolution, frame interpolation, denoising, restoration, and SDR‑to‑HDR conversion to deliver immersive, high‑quality viewing experiences across diverse devices.
AI video processingHDR conversioncloud transcoding
0 likes · 24 min read

The article details the development of '质臻影音' technology by Kuaishou's audio-visual team, which uses AI algorithms to restore classic films and enhance video quality through techniques like super-resolution, noise reduction, and adaptive processing strategies.
AI in video enhancementKuaishou technologyadaptive processing
0 likes · 9 min read


Code DAODec 23, 2021 · Artificial Intelligence
The article explains how satellite images suffer from limited spatial resolution, why the ordering of multi‑temporal frames is irrelevant, and how the PIUnet model introduces permutation‑invariant equivariant layers to achieve state‑of‑the‑art super‑resolution efficiently, winning the AI4EO challenge.
PIUnetSatellite Imagerydeep learning
0 likes · 6 min read

iQiyi's ZoomAI AI-based video restoration and enhancement system won the 2021 CCF Science and Technology Outstanding Progress Award, marking its fifth consecutive CCF honor; the system uses proprietary AI algorithms for super‑resolution, frame interpolation, scratch repair, colorization, etc., restoring classic films and improving viewer experience.
AI video restorationZoomAIartificial intelligence
0 likes · 5 min read

This article presents an overview of Baidu's video foundation technology architecture, covering the video R&D platform, core AI techniques for video understanding, editing, surveillance, and general vision, and detailing innovations such as Attention‑Cluster networks, cross‑modality attention with graph convolution, GANs, super‑resolution, and adaptive encoding.
Adaptive EncodingAttention MechanismGAN
0 likes · 14 min read

This article introduces ImageDT's PlugNet, which combines deep‑learning OCR and super‑resolution techniques to improve low‑quality text recognition in natural scenes, detailing the company's background, OCR challenges, deep‑learning approaches, super‑resolution methods, the PlugNet architecture, experimental results, and future research directions.
AILow-Quality TextOCR
0 likes · 16 min read

AI-driven super-resolution, once limited to academic research, is now tackling real-time video communication challenges by evolving from early interpolation methods to deep learning models, addressing issues of model size, generalization, and real-world degradation, while lightweight networks and encoding-aware techniques promise practical deployment.
AIimage enhancementreal-time video
0 likes · 12 min read

This article explains how Baidu App leverages a VDSR‑based deep‑learning model and a series of mobile‑side optimizations to deliver real‑time image and video super‑resolution on iOS and Android devices, detailing the technical challenges, performance gains, and integration steps.
AIAndroiddeep learning
0 likes · 6 min read

The article describes Baidu App's real-time mobile super-resolution using a VDSR-based model with pruning and depthwise separable convolutions, optimized via application-layer and inference engine techniques to halve latency and memory, enabling on-device high‑def image/video enhancement, reducing server load, and supporting iOS/Android integration.
image enhancementmobile AIreal-time processing
0 likes · 8 min read

This guide shows how to restore and enhance century‑old black‑and‑white Beijing footage using three open‑source AI tools—DAIN for frame interpolation, ESRGAN for super‑resolution, and DeOldify for colorization—complete with setup steps, code snippets, and usage instructions.
AI video restorationDAINDeOldify
0 likes · 10 min read

Anime4K is a GPU‑accelerated super‑resolution algorithm tailored for animated images that achieves single‑digit millisecond latency, and the article explains its underlying residual‑based principle, practical results, and performance limitations on both GPU and CPU platforms.
Anime4KGPUReal-time
0 likes · 3 min read

Anime4K is a GPU-accelerated super‑resolution algorithm designed for animated images, achieving sub‑10 ms latency by enhancing low‑resolution frames with edge‑sharpened residuals, and the article details its principles, visual results, and real‑time performance limitations on typical CPUs.
Anime4KGPUReal-time
0 likes · 2 min read

iQIYI’s ZoomAI combines super‑resolution, denoising, sharpening and color‑enhancement AI models to restore low‑quality historic Chinese films, achieving up to 500‑fold faster processing than manual methods while preserving original visual characteristics.
AI video enhancementZoomAIiQIYI
0 likes · 9 min read

Jiang Zidong explains iQIYI's ZoomAI video enhancement tech, covering super‑resolution, denoising/sharpening, color correction, scratch removal, and frame interpolation, and its modular deployment across business lines for restoring classic content and boosting low‑resolution media, achieving massive efficiency gains.
AIDenoisingZoomAI
0 likes · 29 min read

The YOUKU‑VSRE 2019 challenge invites researchers to develop state‑of‑the‑art video enhancement and super‑resolution models using the largest, most diverse simulated‑noise dataset, with three competition stages (preliminary, semi‑final, final), cash prizes up to ¥100,000, certificates, and fast‑track recruitment opportunities at Alibaba (Youku).
AI challengeCompetitionComputer Vision
0 likes · 3 min read

This tutorial outlines a curriculum covering open‑source examples for enhancing image resolution using SRGAN, explains GAN‑based super‑resolution concepts, details network architectures and perceptual loss, and provides a simple MNIST training walkthrough with code links and resources.
GANMNISTSRGAN
0 likes · 7 min read

Youku’s new video face‑swapping service lets users replace a celebrity’s face with a single uploaded photo by employing a 3D generative model, deep‑learning segmentation, multi‑scale super‑resolution, and trajectory smoothing to achieve fast, near‑photorealistic results across varied angles, expressions, and lighting, though it still lacks personalized models and struggles with extreme side views or heavy occlusions.
3D modelingAIdeepfake
0 likes · 10 min read

ZoomAI is an AI‑driven video enhancement platform that combines modular super‑resolution, denoising, sharpening, color correction, and scratch removal techniques, offering both cloud and mobile SDK solutions for restoring old footage, improving streaming content, and boosting visual quality across devices.
AI video enhancementDenoisingOpenGL
0 likes · 9 min read
