本仓库聚合 2024–2025 顶会/顶刊 与本项目密切相关的实现,均为真克隆(非 submodule),并配套训练/可视化/评测工具。
iqa_methods/:图像/多视图质量评估(IQA)backbones/:视觉骨架/低层复原主干(ViT/Mamba/CNN/KAN)other_methods/:深度/立体/光场/超分/去噪/去雾tools/:训练可视化、画图、特征可视化、数据/MOS 预处理、排序损失与评测协议eval_compare/:IQA 指标库整合 + 一键对比/消融脚本MOS_ref − MOS_i);可选组内 z-score 仅用于训练稳性(汇报用原始 MOS)1 − PLCC + L1 + 组内 list-wise 排序 (+ L_delta);做“同参数一致性”再加 L_param_align| 名称 | 做什么 | Venue/Year | 论文 | 上游代码 | 本地路径 |
|---|---|---|---|---|---|
| LoDa | ViT 高效适配 + 局部增强的 NR-IQA | CVPR 2024 | CVF: https://openaccess.thecvf.com/content/CVPR2024/papers/Xu_Boosting_Image_Quality_Assessment_through_Efficient_Transformer_Adaptation_with_Local_CVPR_2024_paper.pdf | https://github.com/NeosXu/LoDa | iqa_methods/LoDa |
| CrossScore | 多视图 cross-attention 相互参照 | ECCV 2024 | ECCV: https://eccv.ecva.net/virtual/2024/poster/1491/ | https://github.com/ActiveVisionLab/CrossScore | iqa_methods/CrossScore |
| QCN | 几何顺序学习的 BIQA | CVPR 2024 | CVF: https://openaccess.thecvf.com/content/CVPR2024/papers/Shin_Blind_Image_Quality_Assessment_Based_on_Geometric_Order_Learning_CVPR_2024_paper.pdf | https://github.com/nhshin-mcl/QCN | iqa_methods/QCN |
| SaTQA | 监督对比 + Transformer 的 NR-IQA | AAAI 2024 | AAAI: https://ojs.aaai.org/index.php/AAAI/article/view/28285 | https://github.com/Srache/SaTQA | iqa_methods/SaTQA |
可选(若
SKIP_OPTIONAL=0):DepictQA(ECCV’24)、FineVQ(CVPR’25, VQA)、SAMA(AAAI’24)
| 名称 | 做什么 | Venue/Year | 论文/说明 | 上游代码 | 本地路径 |
|---|---|---|---|---|---|
| Fast-iTPN | 分层 ViT + Token Migration | TPAMI 2024 | 见仓库 | https://github.com/sunsmarterjie/iTPN | backbones/iTPN |
| VMamba | 视觉 SSM(线性复杂度) | NeurIPS 2024 | Proc | https://github.com/MzeroMiko/VMamba | backbones/VMamba |
| ViM | Vision Mamba 变体 | ICML 2024 | 见仓库 | https://github.com/hustvl/Vim | backbones/Vim |
| 2DMamba | 2D 选择性扫描 Mamba | CVPR 2025 | 见仓库 | https://github.com/AtlasAnalyticsLab/2DMamba | backbones/2DMamba |
| MambaIR | 面向复原的 Mamba 框架 | ECCV 2024 | 见仓库 | https://github.com/csguoh/MambaIR | backbones/MambaIR |
| EVSSM | VSSM 去模糊骨干/框架 | CVPR 2025 | 见仓库 | https://github.com/kkkls/EVSSM | backbones/EVSSM |
| MS-VMamba | VMamba 多尺度改进 | NeurIPS 2024 | 见仓库 | https://github.com/YuHengsss/MSVMamba | backbones/MSVMamba |
| ConvNeXt-V2 | 强力 CNN 基线组件 | — | 见仓库 | https://github.com/facebookresearch/ConvNeXt-V2 | backbones/ConvNeXt-V2 |
| timm | 预训练模型/模块库 | — | 见仓库 | https://github.com/rwightman/pytorch-image-models | backbones/timm |
| pykan | KAN 模块 | ICLR 2024 | 见仓库 | https://github.com/KindXiaoming/pykan | backbones/pykan |
other_methods/depth_stereo/*(Depth-Anything-V2、Video-Depth-Anything、DAC、DepthPro、Selective-Stereo、TC-Stereo、BiDAStereo、Stereo-Anywhere、Murre、Metric3D、FoundationStereo)other_methods/light_field/*(NeLF-Pro、PDistgNet、PSWPP、BasicLFSR)other_methods/super_resolution/*(PFT-SR、PiSA-SR、PASD、StableVSR、InvSR)other_methods/denoising/*(SplitterNet、RDDM、DualDn、TAP)other_methods/weather_removal/*(Histoformer、MWFormer、MetaWeather、DCL、DarkIR)各条目的论文链接与上游地址,见子目录 SOURCE_URL.txt 或上游仓库 README。
| 名称 | 类别 | 说明 | 上游代码 | 本地路径 |
|---|---|---|---|---|
| pytorch-grad-cam | 特征可视化 | 支持 CNN/ViT 的 CAM 可视 | https://github.com/jacobgil/pytorch-grad-cam | tools/vis/pytorch-grad-cam |
| torch-cam | 特征可视化 | 钩子机制 CAM 工具 | https://github.com/frgfm/torch-cam | tools/vis/torch-cam |
| tensorboardX | 训练可视化 | 标量/图像/网络/Embedding | https://github.com/lanpa/tensorboardX | tools/train/tensorboardX |
| Albumentations | 数据增强 | 高性能 CV 增强 | https://github.com/albumentations-team/albumentations | tools/data/albumentations |
| local_utils | 本仓自带 | mos_utils.py(z-score/ΔMOS/4/5PL/PLCC/SRCC/KRCC)、plotting_utils.py(训练曲线/IQA散点+拟合)、rank_losses.py(ListNet/Pairwise)、data_utils.py(GroupBalancedBatchSampler)、eval_protocol.py(统一评测) | — | tools/local_utils/ |
| 名称 | 说明 | 上游代码 | 本地路径 |
|---|---|---|---|
| PIQ | PyTorch Image Quality:SSIM/PSNR/FSIM/LPIPS 等 | https://github.com/photosynthesis-team/piq | eval_compare/piq |
| LPIPS | Learned Perceptual Image Patch Similarity | https://github.com/richzhang/PerceptualSimilarity | eval_compare/LPIPS |
| pyiqa | IQA 工具箱(FR/NR) | https://github.com/chaofengc/IQA-PyTorch | eval_compare/pyiqa |
| run_eval.py | 一键读取 pred.csv/gt.csv → 4/5PL 标定 → PLCC/SRCC/KRCC/RMSE | (本仓脚本) | eval_compare/run_eval.py |
rg -n -e 'def main\(|if __name__ == .__main__|argparse'rg -n -e 'class .*Dataset|DataLoader|__getitem__|__len__'rg -n -e 'list.?wise|rank|PLCC|SRCC|KRCC|logistic'rg -n -e 'stereo|cost volume|disparit|depth|mvs' other_methods/depth_stereorg -n 'tools/local_utils'本仓库聚合 2024–2025 顶会/顶刊 与本项目密切相关的实现,均为真克隆(非 submodule),并配套训练/可视化/评测工具。
iqa_methods/:图像/多视图质量评估(IQA)backbones/:视觉骨架/低层复原主干(ViT/Mamba/CNN/KAN)other_methods/:深度/立体/光场/超分/去噪/去雾tools/:训练可视化、画图、特征可视化、数据/MOS 预处理、排序损失与评测协议eval_compare/:IQA 指标库整合 + 一键对比/消融脚本MOS_ref − MOS_i);可选组内 z-score 仅用于训练稳性(汇报用原始 MOS)1 − PLCC + L1 + 组内 list-wise 排序 (+ L_delta);做“同参数一致性”再加 L_param_align| 名称 | 做什么 | Venue/Year | 论文 | 上游代码 | 本地路径 |
|---|---|---|---|---|---|
| LoDa | ViT 高效适配 + 局部增强的 NR-IQA | CVPR 2024 | CVF: https://openaccess.thecvf.com/content/CVPR2024/papers/Xu_Boosting_Image_Quality_Assessment_through_Efficient_Transformer_Adaptation_with_Local_CVPR_2024_paper.pdf | https://github.com/NeosXu/LoDa | iqa_methods/LoDa |
| CrossScore | 多视图 cross-attention 相互参照 | ECCV 2024 | ECCV: https://eccv.ecva.net/virtual/2024/poster/1491/ | https://github.com/ActiveVisionLab/CrossScore | iqa_methods/CrossScore |
| QCN | 几何顺序学习的 BIQA | CVPR 2024 | CVF: https://openaccess.thecvf.com/content/CVPR2024/papers/Shin_Blind_Image_Quality_Assessment_Based_on_Geometric_Order_Learning_CVPR_2024_paper.pdf | https://github.com/nhshin-mcl/QCN | iqa_methods/QCN |
| SaTQA | 监督对比 + Transformer 的 NR-IQA | AAAI 2024 | AAAI: https://ojs.aaai.org/index.php/AAAI/article/view/28285 | https://github.com/Srache/SaTQA | iqa_methods/SaTQA |
可选(若
SKIP_OPTIONAL=0):DepictQA(ECCV’24)、FineVQ(CVPR’25, VQA)、SAMA(AAAI’24)
| 名称 | 做什么 | Venue/Year | 论文/说明 | 上游代码 | 本地路径 |
|---|---|---|---|---|---|
| Fast-iTPN | 分层 ViT + Token Migration | TPAMI 2024 | 见仓库 | https://github.com/sunsmarterjie/iTPN | backbones/iTPN |
| VMamba | 视觉 SSM(线性复杂度) | NeurIPS 2024 | Proc | https://github.com/MzeroMiko/VMamba | backbones/VMamba |
| ViM | Vision Mamba 变体 | ICML 2024 | 见仓库 | https://github.com/hustvl/Vim | backbones/Vim |
| 2DMamba | 2D 选择性扫描 Mamba | CVPR 2025 | 见仓库 | https://github.com/AtlasAnalyticsLab/2DMamba | backbones/2DMamba |
| MambaIR | 面向复原的 Mamba 框架 | ECCV 2024 | 见仓库 | https://github.com/csguoh/MambaIR | backbones/MambaIR |
| EVSSM | VSSM 去模糊骨干/框架 | CVPR 2025 | 见仓库 | https://github.com/kkkls/EVSSM | backbones/EVSSM |
| MS-VMamba | VMamba 多尺度改进 | NeurIPS 2024 | 见仓库 | https://github.com/YuHengsss/MSVMamba | backbones/MSVMamba |
| ConvNeXt-V2 | 强力 CNN 基线组件 | — | 见仓库 | https://github.com/facebookresearch/ConvNeXt-V2 | backbones/ConvNeXt-V2 |
| timm | 预训练模型/模块库 | — | 见仓库 | https://github.com/rwightman/pytorch-image-models | backbones/timm |
| pykan | KAN 模块 | ICLR 2024 | 见仓库 | https://github.com/KindXiaoming/pykan | backbones/pykan |
other_methods/depth_stereo/*(Depth-Anything-V2、Video-Depth-Anything、DAC、DepthPro、Selective-Stereo、TC-Stereo、BiDAStereo、Stereo-Anywhere、Murre、Metric3D、FoundationStereo)other_methods/light_field/*(NeLF-Pro、PDistgNet、PSWPP、BasicLFSR)other_methods/super_resolution/*(PFT-SR、PiSA-SR、PASD、StableVSR、InvSR)other_methods/denoising/*(SplitterNet、RDDM、DualDn、TAP)other_methods/weather_removal/*(Histoformer、MWFormer、MetaWeather、DCL、DarkIR)各条目的论文链接与上游地址,见子目录 SOURCE_URL.txt 或上游仓库 README。
| 名称 | 类别 | 说明 | 上游代码 | 本地路径 |
|---|---|---|---|---|
| pytorch-grad-cam | 特征可视化 | 支持 CNN/ViT 的 CAM 可视 | https://github.com/jacobgil/pytorch-grad-cam | tools/vis/pytorch-grad-cam |
| torch-cam | 特征可视化 | 钩子机制 CAM 工具 | https://github.com/frgfm/torch-cam | tools/vis/torch-cam |
| tensorboardX | 训练可视化 | 标量/图像/网络/Embedding | https://github.com/lanpa/tensorboardX | tools/train/tensorboardX |
| Albumentations | 数据增强 | 高性能 CV 增强 | https://github.com/albumentations-team/albumentations | tools/data/albumentations |
| local_utils | 本仓自带 | mos_utils.py(z-score/ΔMOS/4/5PL/PLCC/SRCC/KRCC)、plotting_utils.py(训练曲线/IQA散点+拟合)、rank_losses.py(ListNet/Pairwise)、data_utils.py(GroupBalancedBatchSampler)、eval_protocol.py(统一评测) | — | tools/local_utils/ |
| 名称 | 说明 | 上游代码 | 本地路径 |
|---|---|---|---|
| PIQ | PyTorch Image Quality:SSIM/PSNR/FSIM/LPIPS 等 | https://github.com/photosynthesis-team/piq | eval_compare/piq |
| LPIPS | Learned Perceptual Image Patch Similarity | https://github.com/richzhang/PerceptualSimilarity | eval_compare/LPIPS |
| pyiqa | IQA 工具箱(FR/NR) | https://github.com/chaofengc/IQA-PyTorch | eval_compare/pyiqa |
| run_eval.py | 一键读取 pred.csv/gt.csv → 4/5PL 标定 → PLCC/SRCC/KRCC/RMSE | (本仓脚本) | eval_compare/run_eval.py |
rg -n -e 'def main\(|if __name__ == .__main__|argparse'rg -n -e 'class .*Dataset|DataLoader|__getitem__|__len__'rg -n -e 'list.?wise|rank|PLCC|SRCC|KRCC|logistic'rg -n -e 'stereo|cost volume|disparit|depth|mvs' other_methods/depth_stereorg -n 'tools/local_utils'