没有合适的资源?快使用搜索试试~ 我知道了~
(OpenCV CG匹配算法)Realistic CG Stereo Image Dataset with Ground Tru
需积分: 0 0 下载量 90 浏览量
2022-08-04
00:18:41
上传
评论
收藏 365KB PDF 举报
温馨提示
试读
4页
Realistic CG Stereo Image Dataset with Ground Truth Disparity MapsUniversity of
资源推荐
资源详情
资源评论
Realistic CG Stereo Image Dataset with Ground Truth Disparity Maps
Sara Martull
University of Tsukuba, Japan
E-mail: info@martull.com
Martin Peris
Cyberdyne Inc., Japan
E-mail: martin
peris@cyberdyne.jp
Kazuhiro Fukui
University of Tsukuba, Japan
E-mail: [email protected].ac.jp
Abstract
Stereo matching is one of the most active research
areas in computer vision. While a large number of al-
gorithms for stereo correspondence have been devel-
oped, research in some branches of the field has been
constrained due to the few number of stereo datasets
with ground truth disparity maps available. Having
available a large dataset of stereo images with ground
truth disparity maps would boost the research on new
stereo matching methods, for example, methods based
on machine learning. In this work we develop a large
stereo dataset with ground truth disparity maps using
highly realistic computer graphic techniques. We also
apply some of the most common stereo matching tech-
niques to our dataset to confirm that our highly realistic
CG stereo images remain as challenging as real-world
stereo images. This dataset will also be of great use
for camera tracking algorithms, because we provide the
exact camera position and rotation in every frame.
1. Introduction
Stereo vision is a very active research topic, every
year several new stereo matching methods are intro-
duced [5]. The goal of these stereo matching algorithms
is to accurately generate a dense disparity map, which
describes the difference in location of corresponding
features seen by the left and right cameras. To measure
and compare the performance of such algorithms it is
essential that the ground truth disparity map is known.
Several stereo datasets with known ground truth dis-
parity maps are available [5, 6], but the number of stereo
pairs is very limited. This limitation has been constrain-
ing the progress of research in some branches of the
field. For example, trying to apply machine learning
Figure 1: Head and Lamp scene. Left image and ground
truth disparity map.
(ML) techniques to solve the stereo matching problem
has been very difficult due to the fact that ML usually
requires large amounts of data with ground truth for
learning and very few are available. Some efforts on us-
ing Computer Graphics (CG) synthetic data have been
made [1], but the simplicity of the generated scenes
makes the stereo matching problem unrealistically easy
to solve.
Among the available datasets, one of the most known
and extended scenes is the Head and Lamp (Figure 1)
stereo dataset developed at University of Tsukuba [3].
In this work we created a highly realistic CG dataset
that properly models real-world imperfections, while
providing accurate ground truth. It is based in the orig-
inal Head and Lamp set of images, as a tribute to the
early efforts of the University of Tsukuba in stereo vi-
sion, giving the chance to appreciate the scene from
points of view not seen until now.
In addition, this dataset will be very useful for cam-
era tracking algorithms, since we can provide the 3D
position and rotation of the camera in every frame of
the sequence.
Since we are working in a 3D environment, we can
create any possible camera setting, and modify any
camera parameter just as we would do with a real-world
camera. In future datasets we will include video se-
资源评论
点墨楼
- 粉丝: 32
- 资源: 279
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功