HM-TIR
收藏资源简介:
HM-TIR数据集包含清洁和退化红外图像,数据集结构包括:imgs(所有清洁红外图像)、composited(测试复合退化红外图像,包含normal和hard子集)、single(单一退化红外图像,包含blur、contrast和noise子集)、test(测试清洁红外图像)和meta(元数据,包括train.txt和test.txt)
The HM-TIR dataset contains clean and degraded infrared images. Its structure includes the following components: imgs (all clean infrared images), composited (test composite degraded infrared images with normal and hard subsets), single (single-degradation infrared images with blur, contrast and noise subsets), test (test clean infrared images), and meta (metadata including train.txt and test.txt).
HM-TIR 数据集概述
数据集基本信息
- 数据集名称:HM-TIR
- 研究主题:红外图像增强
- 相关论文:Enhancing Infrared Vision: Progressive Prompt Fusion Network and Benchmark (NeurIPS 2025)
- 论文链接:https://arxiv.org/abs/2510.09343
数据集内容
图像类型
- 干净红外图像
- 退化红外图像
数据集结构
HM-TIR ├── imgs (所有干净红外图像) ├── composited (复合退化红外图像测试集) │ ├── normal (正常退化子集) │ └── hard (困难退化子集) ├── single (单一退化红外图像) │ ├── blur (模糊退化子集) │ ├── contrast (对比度退化子集) │ └── noise (噪声退化子集) ├── test (测试用干净红外图像) └── meta (元数据,包含train.txt, test.txt)
数据集获取
- 下载地址:https://drive.google.com/file/d/17kq2NQ-8bFeESlanffcOSlFLwGb0Y8Rq/view?usp=sharing
应用场景
- 红外图像增强研究
- 退化图像恢复
- 图像质量评估
相关资源
- 预训练模型:https://drive.google.com/file/d/1Yt_YViFCpxgO-MfBlINVt4-jHAgsalv_/view?usp=sharing
- 真实世界测试集:https://drive.google.com/file/d/1uhWCvPExarnUL-r_G0MW530RXKpj5izJ/view?usp=sharing
- 结果示例:https://drive.google.com/file/d/10jiNGF0c6H0cR83yEN5I3stLf9qiX6cR/view?usp=sharing
联系方式
- Zihang Chen: chenzi_hang@mail.dlut.edu.cn
- Jinyuan Liu: atlantis918@hotmail.com



