HAND Dataset
==== =======
A. Mittal, A. Zisserman and P. H. S. Torr
Introduction
------------
We introduce a comprehensive dataset of hand images collected from various different public image data set sources as listed in [1]. A total of 13050 hand instances are annotated. Hand instances larger than a fixed area of bounding box (1500 sq. pixels) are considered 'big' enough for detections and are used for evaluation. This gives around 4170 high quality hand instances. While collecting the data, no restriction was imposed on the pose or visibility of people, nor was any constraint imposed on the environment.
In each image, all the hands that can be perceived clearly by humans are annotated. The annotations consist of a bounding rectangle, which does not have to be axis aligned, oriented with respect to the wrist. Please cite [1] if you use this dataset.
Contents
--------
This package contains:
- Training dataset (all the images with hand bounding-box annotations).
- Code to visualize annotations.
- Code to give statistics of data in the dataset.
Let PWD be the directory of the hand-dataset, then the structure of the containts is as follows:
PWD/training_data: Directory containing training data.
PWD/training_data/images: Directory containing training data images.
PWD/training_data/annotations: Annotations files for the training dataset.
counttraindata.m: Matlab function giving training data statistics.
displaydata.m : Matlab function to visualize annotations over images.
Description of the annotation files
----------- -- --- ---------- -----
The annotation files are in standard matlab '.mat' format which contain annotations for the four end-points of the hand bounding box.
It has a structure 'boxes' having hand-boxes as different indices of cell array. Please refer to the visualization function (displaydata.m) for more clearification.
Evaluation Criteria
---------- --------
The performance is evaluated using average precision (AP) (the area under the Precision Recall curve). A hand detection is considered true or false according to its overlap with the ground-truth bounding box.
A box is positive if the overlap score is more than 0.5. The overlap ratio is computed between the axis-aligned bounding rectangles around the ground-truth and the detected hand bounding box.
Support
-------
For any query/suggestions, please drop an email to the following addresses:
arpit@robots.ox.ac.uk
az@robots.ox.ac.uk
philiptorr@brookes.ac.uk
References
----------
[1] Hand detection using multiple proposals
A. Mittal, A. Zisserman, P.H.S. Torr
Proceedings of British Machine Vision Conference, 2011.
没有合适的资源?快使用搜索试试~ 我知道了~
人手检测训练数据库
共2000个文件
mat:2964个
ds_store:6个
m:4个
3星 · 超过75%的资源 需积分: 50 184 下载量 35 浏览量
2012-12-05
10:32:21
上传
评论
收藏 1.52MB RAR 举报
温馨提示
人手检测系统的训练图像集,其中Matlab文件中详细标注了人手出现的位置和大小,训练后的系统可以很好的检测出人手
资源详情
资源评论
资源推荐
收起资源包目录
人手检测训练数据库 (2000个子文件)
.DS_Store 363KB
.DS_Store 6KB
.DS_Store 6KB
._.DS_Store 82B
._.DS_Store 82B
._.DS_Store 82B
counttraindata.m 3KB
displaydata.m 781B
._displaydata.m 82B
._counttraindata.m 82B
Skin_8.mat 2KB
Inria_263.mat 2KB
Poselet_355.mat 1KB
VOC2007_434.mat 1KB
Poselet_356.mat 1KB
Poselet_98.mat 1KB
Poselet_124.mat 1KB
Poselet_366.mat 1KB
Poselet_20.mat 1KB
VOC2007_568.mat 1KB
Poselet_318.mat 1KB
VOC2007_244.mat 1KB
Poselet_37.mat 1KB
Inria_261.mat 1KB
Inria_91.mat 1KB
Poselet_349.mat 1019B
Poselet_170.mat 1014B
Inria_260.mat 1013B
VOC2007_165.mat 1011B
Inria_258.mat 997B
Poselet_11.mat 988B
Poselet_314.mat 948B
VOC2007_50.mat 943B
VOC2007_714.mat 943B
Inria_257.mat 937B
Poselet_5.mat 936B
Poselet_4.mat 935B
Inria_556.mat 931B
VOC2007_535.mat 929B
Inria_259.mat 910B
Inria_175.mat 904B
Poselet_54.mat 899B
Skin_95.mat 878B
Poselet_368.mat 873B
Poselet_38.mat 872B
Poselet_19.mat 871B
VOC2010_1032.mat 871B
VOC2007_525.mat 870B
Poselet_243.mat 870B
VOC2007_42.mat 870B
VOC2010_1052.mat 869B
VOC2010_1000.mat 869B
Skin_108.mat 868B
Poselet_67.mat 868B
Poselet_63.mat 865B
VOC2007_33.mat 863B
Poselet_350.mat 861B
Poselet_82.mat 837B
Poselet_357.mat 835B
Skin_271.mat 832B
Inria_415.mat 827B
Poselet_310.mat 805B
Inria_90.mat 801B
Skin_82.mat 800B
Inria_545.mat 799B
Poselet_352.mat 799B
Skin_290.mat 798B
Inria_498.mat 798B
Poselet_317.mat 798B
VOC2010_1295.mat 797B
Skin_230.mat 796B
Poselet_150.mat 796B
Skin_228.mat 796B
Poselet_139.mat 795B
Poselet_191.mat 795B
Inria_375.mat 795B
Poselet_58.mat 795B
VOC2007_52.mat 794B
Skin_126.mat 794B
VOC2007_509.mat 793B
VOC2010_1301.mat 793B
VOC2007_698.mat 791B
Poselet_79.mat 790B
VOC2007_172.mat 789B
Poselet_6.mat 787B
Inria_67.mat 785B
Poselet_65.mat 764B
Poselet_62.mat 764B
VOC2007_452.mat 761B
Skin_166.mat 758B
Inria_264.mat 732B
Inria_491.mat 727B
Poselet_57.mat 725B
VOC2010_110.mat 725B
Poselet_324.mat 723B
VOC2007_735.mat 723B
Poselet_276.mat 723B
Skin_55.mat 723B
Poselet_334.mat 723B
Poselet_51.mat 723B
共 2000 条
- 1
- 2
- 3
- 4
- 5
- 6
- 20
bushukukei123
- 粉丝: 0
- 资源: 3
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
评论4