转载请注明作者和出处:http://blog.csdn.net/john_bh/
Depth super resolution 论文及代码汇总,持续更新中~~
文章目录
1. Dataset
- 合成数据:IRS
SCS: Stereo Camera Sensing
| DataSet | Ref. | Year | Scene | Sensor Type | Sensor Name | Data Modalities | Size | Extra Data | Introduction | Website |
|---|---|---|---|---|---|---|---|---|---|---|
| Middlebury 2001 | - | 2001 | Indoor | SCS | Stereo Camera | Color, Depth | 6 Images | - | - | Website |
| Middlebury 2003 | - | 2003 | Indoor | Structured Light | Custom-build, Structured Light |
Color, Depth | 2 Images | - | F-full size: 1800 x 1500 H-half size: 900 x 750 Q-quarter size: 450 x 375 |
Website |
| Middlebury 2005 | - | 2005 | Indoor | Structured Light | Custom-build, Structured Light |
Color, Depth | 9 Images(3 GT hidden) | - | F-full size: 1330…1390 x 1110 H-half size: 665…695x 555 T-third size: 443…463x 370 |
Website |
| Middlebury 2006 | CVPR2007 | 2006 | Indoor | Structured Light | Custom-build, Structured Light |
Color, Depth | 21 Images | - | F-full size: 1240…1396x 1110 H-half size: 620…698x 555 T-third size: 413…465x 370 |
Website |
| Middlebury 2014 | - | 2014 | Indoor | SCS | Stereo Camera | Color, Depth | 33 Images(10 GT hidden) | - | F-full size: 2300…3000x1848…2020 | Website |
| Middlebury 2021 | - | 2021 | Indoor | Structured Light | Custom-build, Structured Light |
Color, Depth | 24 scenes | - | F-full size: 1920x1080 ladder1,ladder2:1080x1920 |
Website |
| NYUv2 | - | 2012 | Indoor | Structured Light | Kinect V1 | Color, Depth,Accelerometer | 464 Scenes | Semantic Segmentation | 640x480 407024 Frames With 1449 Labeled Aligned RGB-D Images |
Official Website2 |
| Lu | CVPR2014 | 2014 | Indoor | Structured Light | ASUS Xtion Pro | Color, Depth | 6 Scenes | - | 640x480 | Website |
| Lu_Middlebury | - | 2014 | Indoor | - | - | Color, Depth | 30 Images | - | 417…465x370…381 The ground truth color and depth images are from the Middlebury dataset [2001, 2003, 2005, 2006]. Here are the 30 images used in Lu paper |
Website |
| RGBD-D | CVPR2021 | 2021 | Indoor, Outdoor |
Time of Flight | Huawei P30 Pro(LR), Helios(GT) |
Color, Depth | 4811 paired Scenes | - | HR RGBD(GT): 512X38 LR Depth(fill):192x144 Raw LR depth:240x180 Raw HR depth:640x480 2215 for train,405 for test |
Website |
| DIML | paper | 2021 | Indoor | SCS,ToF | Kinectv2(Indoor) | Color, Depth | 2000 Indoor | - | HR:1344x756 LR: 512x288 filled depth images using colorization (mm, 16bit png), Raw depth images (mm, 16bit png), Normalized depth images for visualization purpose (8bit png), Color images obtained from Kinect v2 (8bit png) |
Website |
| DIML | paper | 2021 | Outdoor | SCS,ToF | ZED Stereo Camera(Outdoor) |

本文汇总了深度超分辨率领域的最新论文和代码,涵盖多个研究方向和年度,包括2023年的一篇深度多尺度特征融合方法,以及2024年的一篇学习连续深度表示的DSR方法。这些方法利用RGB信息和深度图结构,通过多阶段网络和注意力机制提升深度图的分辨率。实验表明,提出的框架在多个基准上表现优秀,特别是在高倍数上采样时,优于现有技术。

3051

被折叠的 条评论
为什么被折叠?



