ntire

"Retinexformer: One-stage Retinex-based Transformer for Low-light Image Enhancement" (ICCV 2023 Top-10 Cited 🏆) & (NTIRE 2024 Runner-Up 🏆) & (NTIRE 2025 Winner 🏆) & (NTIRE 2026 Winner 🏆)

A toolbox for spectral compressive imaging reconstruction including MST (CVPR 2022), CST (ECCV 2022), DAUHST (NeurIPS 2022), BiSCI (NeurIPS 2023), HDNet (CVPR 2022), MST++ (CVPRW 2022), etc.

[ECCV] Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration. Advances in Image Manipulation (AIM) workshop ECCV 2022. Try it out! over 3.3M runs https://replicate.com/mv-lab/swin2sr

"MST++: Multi-stage Spectral-wise Transformer for Efficient Spectral Reconstruction" (CVPRW 2022) & (Winner of NTIRE 2022 Spectral Recovery Challenge) and a toolbox for spectral reconstruction

Official PyTorch implementation of DW-GAN, 1st place solution of NTIRE 2021 NonHomogeneous Dehazing Challenge (CVPR Workshop 2021).

PyTorch Implementation of "Densely Connected Hierarchical Network for Image Denoising", CVPRW, NTIRE2019

XReflection is a neat toolbox tailored for single-image reflection removal(SIRR). We offer state-of-the-art SIRR solutions for training and inference, with a high-performance data pipeline, multi-GPU/TPU/NPU support, and more!

Solution for NTIRE2018 Image Dehazing Challenge & ACCV2018 Kangfu Mei et al.

Official PyTorch implementation of dehazing method based on FFC and ConvNeXt, 1st place solution of NTIRE 2023 HR NonHomogeneous Dehazing Challenge (CVPR Workshop 2023).

Series solutions for Faster SR (NTIRE/ACCV)

Pytorch implementation of "Multi-scale Single Image Dehazing using Perceptual Pyramid Deep Network"

The 1st place solution to CVPR 2021 NTIRE Depth Guided Image One-to-One Relighting MBNet

NTIRE 2021, Image Quality Assessment Challenge (MACS team)

Efficient Space-time Super Resolution using Flow and Mask Upsampling

The 1st place solution to CVPR2021 NTIRE Depth Guided Relighting Challenge Track 2: Any-to-any relighting

Generalized AI-generated image detector on the NTIRE 2026 in-the-wild benchmark, tested on unseen generators like Nano Banana (DINOv3 + LoRA).

Multi-scale features + parallel transformers for full-reference image quality assessment (arXiv:2204.09779, NTIRE 2022 @ CVPRW)

Content-variant reference IQA via knowledge distillation: a full-reference teacher transfers high-quality distribution priors to a student that needs only a non-pixel-aligned / content-variant reference, so quality scores no longer require a pixel-perfect reference image.

Enhanced version of B2SCVR prepared under B.Tech Mini Project and NTIRE competition