voc-release
所属分类:其他
开发工具:matlab
文件大小:1185KB
下载次数:9
上传日期:2017-09-16 19:54:12
上 传 者:
云墨染
说明: 可以用来识别视频中的人数,识别物体的数目。
(Can be used to identify the number of people in the video, the number of objects identified.)
文件列表:
voc-release1\clipboxes.m (296, 2008-05-10)
voc-release1\compile.m (52, 2008-05-10)
voc-release1\COPYING (1079, 2008-05-10)
voc-release1\demo.m (812, 2008-05-10)
voc-release1\detect.m (6175, 2008-05-10)
voc-release1\dt.cc (2511, 2008-05-10)
voc-release1\fconv.cc (2558, 2008-05-10)
voc-release1\featpyramid.m (1166, 2008-05-10)
voc-release1\features.cc (5683, 2008-05-13)
voc-release1\foldHOG.m (138, 2008-05-10)
voc-release1\HOGpicture.m (622, 2008-05-10)
voc-release1\image1.jpg (35232, 2008-05-10)
voc-release1\image2.jpg (17311, 2008-05-10)
voc-release1\models (0, 2017-09-15)
voc-release1\models\bicycle_model.mat (91968, 2008-05-10)
voc-release1\models\bicycle_pr.jpg (37347, 2008-05-10)
voc-release1\models\bus_model.mat (99180, 2008-05-10)
voc-release1\models\bus_pr.jpg (36700, 2008-05-10)
voc-release1\models\car_model.mat (104695, 2008-05-10)
voc-release1\models\car_pr.jpg (35979, 2008-05-10)
voc-release1\models\cat_model.mat (104995, 2008-05-10)
voc-release1\models\cat_pr.jpg (37986, 2008-05-10)
voc-release1\models\cow_model.mat (91962, 2008-05-10)
voc-release1\models\cow_pr.jpg (36807, 2008-05-10)
voc-release1\models\dog_model.mat (92508, 2008-05-10)
voc-release1\models\dog_pr.jpg (37483, 2008-05-10)
voc-release1\models\horse_model.mat (92508, 2008-05-10)
voc-release1\models\horse_pr.jpg (37150, 2008-05-10)
voc-release1\models\motorbike_model.mat (95293, 2008-05-10)
voc-release1\models\motorbike_pr.jpg (38027, 2008-05-10)
voc-release1\models\person_model.mat (45421, 2008-05-10)
voc-release1\models\person_pr.jpg (37364, 2008-05-10)
voc-release1\models\scores (173, 2008-05-10)
voc-release1\models\sheep_model.mat (78135, 2008-05-10)
voc-release1\models\sheep_pr.jpg (37107, 2008-05-10)
voc-release1\nms.m (997, 2008-05-10)
voc-release1\resize.cc (3250, 2008-05-10)
voc-release1\showboxes.m (487, 2008-05-10)
voc-release1\visualizeHOG.m (491, 2008-05-10)
... ...
Information
===========
This is an implementation of the system described in:
P. Felzenszwalb, D. McAllester, D. Ramaman.
A Discriminatively Trained, Multiscale, Deformable Part Model.
To appear in CVPR 2008.
We are currently distributing the object detection code with models
trained on the PASCAL VOC 2006 dataset. The learning code will be
made available in the future.
Please note this implementation is not identical to the original one
(as described in the paper). In particular the part displacements are
not constrained to be in a bounded range. This leads to a slight
improvement in performance for some object categories.
Basic Usage
===========
The system is implemented in matlab, with a few helper functions
written in C++ for efficiency reasons.
Basic steps for running the system:
1. Unpack the code to a directory.
2. Start matlab from the directory with the code.
3. Run the 'compile' script.
This will compile the C++ helper functions.
4. Run the 'demo' script.
This will run the object detector on a test image.
The main functions defined by the code are:
boxes = detect(im, model, thresh)
top = nms(boxes, overlap)
boxes = clipboxes(im, boxes)
showboxes(im, boxes)
visualizemodel(model)
Their usage is demonstrated in the 'demo' script.
The directory 'models' contains matlab datafiles with models trained
on the PASCAL VOC 2006 train+val dataset. The score of each model in
the test dataset is recorded in the file 'scores'.
Loading a model datafile into matlab will define a variable 'model'
with the model trained for a particular object category. The value
'model.thresh' defines a threshold that can be used in the 'detect'
function to obtain a high recall rate (at the expense of many false
positives per image).
近期下载者:
相关文件:
收藏者: