jvgemert/apt: Action localization Proposals from dense Trajectories

原作者: [db:作者] 来自: 网络收藏邀请

开源软件名称（OpenSource Name）：

jvgemert/apt

开源软件地址(OpenSource Url)：

https://github.com/jvgemert/apt

开源编程语言(OpenSource Language)：

Python 88.3%

开源软件介绍(OpenSource Introduction)：

APT: Action localization Proposals from dense Trajectories

Python source code written by Jan van Gemert, if you find any bugs or improvements, feel free to let me know. If you find the code useful, please cite our paper: APT: Action localization Proposals from dense Trajectories.

Note that this code may not run directly out of the box. This is because the third party libraries need to be installed, and they may vary depending on your system. For convenience we include all third party code here, although all credit for them is with the original authors, and we highly recommend them.

Requirements:

python 2.7
numpy
h5py (hdf5 data format)
cython
yael ( best: https://gforge.inria.fr/projects/yael/ or see third/yael_v438.tar.gz; Configure yael with ./configure.sh --enable-numpy. And you may have to add the yael install directory to the PYTHONPATH.)
Improved Trajectory code (http://lear.inrialpes.fr/people/wang/improved_trajectories or see folder third/. Note that DenseTrackStab.cpp is adapted to export the trajectory x and y coordinates per frame, and the video statistics (width, height, lenght) are exported. See the folder third/ for the modified version of DenseTrackStab.cpp. And, see tubePy/extractDT.sh for an example script on how to extract them.

Cython:

The APT package contains Python code with some parts in cython. The code also works without cython, albeit slower (slower parts are: clustering, merging, and IoU computations). To take advantage of the cython speedup you have to compile the tubePy/*.pyx files (see the compile2cython.sh script that calls the individual tubePy/compile*.sh scripts).

Organization

This package is organized as follows.

example/            # example video and groundtruth tube
third/              # third party libs 
tubePy/             # APT python code
doExample.sh        # script that will run the example
compile2cython.sh   # script to compile python functions to cython
README.md           # this file

The APT package is developed under linux mint. The third party libs needs to be installed. The third party software is in:

third/
    DenseTrackStab                  # pre-compiled linux mint 17 64bit executable for improved trajectories
    improved_trajectory_APT         # Slightly modified source code for improved trajectories
    yael_v438.tar.gz                # source code for yael

All credit for third party sofware is with the authors. It is best to rely on the original source code which is taken from: yael and Improved Trajectory code.

doExample.sh

The 'doExample.sh' file has all steps explained to compute APT for the included example, the steps it does are:

1- extract the dense trajectories from the video and put in the expected format --> feat/
2- create Action Proposals from dense Trajectories by clustering --> clusts/
3- convenience function to convert proposals clusters of step 2 to tubes --> tubes/
4- compare proposal clusters/tubes to groundTruth, outputs IoU overlap scores in --> IoU/
5- convert proposal clusters/tubes to (0-based) trajectory IDs, so that the trajectories in a proposal can be mapped to Vlad/Fisher-vector/BoW --> trajIDs/

All steps in this package can be done on the internal cluster format (which will be converted to tubes internally each time), or, they can be done on explicitly exported tube proposals. The reason for providing a 'detour' to tube files, is the ability to use tubes in other software. The example includes all steps for the internal cluster format, and also for tube files.

For more details, read doExample.sh.

Input and output

The included example consists of sub folders. The minimum input requirements are:

example/
    gt/             # Tube ground truth (needed for evaluating proposal quality)
    vid/            # Input video

The output is computed with this package, which results in the following folder structure:

example/
    feat/           # dense trajectory features 
    IoU/            # Intersection over Union scores with ground truth for proposals 
    IoU-tubes/      # Intersection over Union scores with ground truth for proposals in tube format
    clusts/         # Cluster results
    trajIDs/        # The trajectory IDs that belong to proposal tubes
    trajIDs-tubes/  # The trajectory IDs that belong to proposal tubes in tube format
    tube/           # Tube proposals

To start, it is best to read and run the 'doExample.sh' file, it has all steps required to compute APT for an included example video: /example/vid/001.avi.

Frequently Asked Questions

- Did the UCF101 scores change?

Yes. We made a mistake in the ground truth. We corrected our ground truth and make it available. We updated the scores in the paper, the numbers are different but the conclusions do not change.

- Do you have pre-computed tubes available?

Yes.

UCF-Sports (trimmed)

150 sports videos in 10 classes, available from http://crcv.ucf.edu/data/UCF_Sports_Action.php

ucfSportsAptTube.tar.gz : The tube proposals
ucfSportsGT.tar.gz : the ground truth tubes

UCF-101 (trimmed)

3,207 youtube videos for action detection of 24 classes, available from http://crcv.ucf.edu/data/UCF101.php

ucf101AptTube.tar.gz : The tube proposals
ucf101GT.tar.gz : the ground truth tubes

MSR Action Dataset II (untrimmed)

54 videos of 3 classes, available from http://research.microsoft.com/en-us/um/people/zliu/actionrecorsrc/

MSRIIAptTube.tar.gz : The tube proposals
MSRIIGT.tar.gz : the ground truth tubes

- Do you also have the 'Tubelet' results?

Yes. The Tubelet action tube proposals come from our CVPR 2014 paper Action Localization by Tubelets from Motion and are computed on MSR-II MSRIITubeletCVPR14.tar.gz and on UCF-Sports ucfSportsTubeletCVPR14.tar.gz.

- What is the tube dataformat?

A tube is a set of frameNRs and bounding boxes.

Specifically, a tube is represented as an array of 5 numbers: [frameNr, minX, minY, maxX, maxY].

- How do I read a tube from disk?

The tubes are stored in hdf5 a platform-independent binary dataformat.

Examples on how to read tubes:

Matlab:

>> hinfo = hdf5info('example/tube/001/nn10T50.hdf5')
>> nrTubes = numel(hinfo.GroupHierarchy.Datasets)
>> tube1 = hdf5read(hinfo.GroupHierarchy.Datasets(1))
>> tube2 = hdf5read(hinfo.GroupHierarchy.Datasets(2))

Python, using the provide TubeList class:

>>> import h5py
>>> from tubePy.tubeList import TubeList
>>> tubes = TubeList()
>>> tubes.readHDF5('example/tube/001/nn10T50.hdf5')
	reading tubes: example/tube/001/nn10T50.hdf5 ; Number of tubes to read: 2147
>>> tube1 = tubes[0]
>>> print tube1
>>> tube2 = tubes[1]

Python, without the TubeList class:

>>> import h5py
>>> inHDF5file = h5py.File('example/tube/001/nn10T50.hdf5', 'r')
>>> nrTubes = len(inHDF5file)
>>> tube1 = inHDF5file["0"]
>>> print tube1[()]
>>> tube2 = inHDF5file["1"]

Other languages like c/c++ also have hdf5 implementations.

- What does the output look like?

isTrimmedVideo=0;

./doExample.sh 
** Step 0: try to compile parts to cython **
Done compiling

** Step 1: Extract trajectories from the video **
run tubePy/extractDT.sh example/vid/001.avi example/feat/001/idt.hdf5
Saved temporary trajectories to: ./localexample/vid/001.avi/DenseTrackStab.txt.gz

** Convert the trajectories to the format as used for DTP **
python ./tubePy/runConvertDenseTraj2fvec.py ./localexample/vid/001.avi/DenseTrackStab.txt.gz example/feat/001/idt.hdf5
	writing to example/feat/001/idt.hdf5
	Information of video size, length: 55, width: 720, height: 404 ; Write 55 720 404 to: "vidinfo"
	Counting nr of features: 10k ; Total: 14337
	Feature: trajecShape, (nrTraj x dim) =  (14337, 40) ; Write to "trajecShape"
	Feature: trajecTubes, (nrTraj x dim) =  (14337, 60) ; Write to "trajecTubes"
	Feature: trajstab, (nrTraj x dim) =  (14337, 30) ; Write to "trajstab"
	Feature: hogstab, (nrTraj x dim) =  (14337, 96) ; Write to "hogstab"
	Feature: hofstab, (nrTraj x dim) =  (14337, 108) ; Write to "hofstab"
	Feature: mbhstab, (nrTraj x dim) =  (14337, 192) ; Write to "mbhstab"

clean up temporary file; rm  ./localexample/vid/001.avi/DenseTrackStab.txt.gz

** Step 2: Create Action Proposals from dense Trajectories by clustering **
python tubePy/runClust.py example/feat/001/idt.hdf5 10 0 example/clusts/001/nn10.hdf5
	adding (possible) missing features to example/clusts/001/nn10.hdf5

** Possible Step 3: Convenience function to export the clusters of step 2 to proposals **
python tubePy/runClust2tube.py example/clusts/001/nn10.hdf5 example/feat/001/idt.hdf5 50 example/tube/001/nn10T50.hdf5
	Write tubes to: example/tube/001/nn10T50.hdf5
	Reading feature file: example/feat/001/idt.hdf5, "trajecShape" size (14337, 40)
	Reading feature file: example/feat/001/idt.hdf5, "vidinfo" size (3,)
	Feature: "hofstab"; Number of proposals: 48 tot: 48
	Feature: "hofstab-hogstab-trajstab-mbhstab"; Number of proposals: 70 tot: 118
	Feature: "hofstab-mbhstab"; Number of proposals: 62 tot: 180
	Feature: "hofstab-trajstab"; Number of proposals: 63 tot: 243
	Feature: "hofstab-trajstab-mbhstab"; Number of proposals: 67 tot: 310
	Feature: "hogstab"; Number of proposals: 71 tot: 381
	Feature: "hogstab-hofstab"; Number of proposals: 70 tot: 451
	Feature: "hogstab-hofstab-mbhstab"; Number of proposals: 69 tot: 520
	Feature: "hogstab-hofstab-trajstab"; Number of proposals: 66 tot: 586
	Feature: "hogstab-mbhstab"; Number of proposals: 71 tot: 657
	Feature: "hogstab-trajstab"; Number of proposals: 72 tot: 729
	Feature: "hogstab-trajstab-mbhstab"; Number of proposals: 70 tot: 799
	Feature: "mbhstab"; Number of proposals: 71 tot: 870
	Feature: "spat"; Number of proposals: 67 tot: 937
	Feature: "spat-hofstab"; Number of proposals: 69 tot: 1006
	Feature: "spat-hofstab-mbhstab"; Number of proposals: 70 tot: 1076
	Feature: "spat-hofstab-trajstab"; Number of proposals: 64 tot: 1140
	Feature: "spat-hofstab-trajstab-mbhstab"; Number of proposals: 72 tot: 1212
	Feature: "spat-hogstab"; Number of proposals: 70 tot: 1282
	Feature: "spat-hogstab-hofstab"; Number of proposals: 75 tot: 1357
	Feature: "spat-hogstab-hofstab-mbhstab"; Number of proposals: 75 tot: 1432
	Feature: "spat-hogstab-hofstab-trajstab"; Number of proposals: 74 tot: 1506
	Feature: "spat-hogstab-hofstab-trajstab-mbhstab"; Number of proposals: 72 tot: 1578
	Feature: "spat-hogstab-mbhstab"; Number of proposals: 72 tot: 1650
	Feature: "spat-hogstab-trajstab"; Number of proposals: 74 tot: 1724
	Feature: "spat-hogstab-trajstab-mbhstab"; Number of proposals: 76 tot: 1800
	Feature: "spat-mbhstab"; Number of proposals: 70 tot: 1870
	Feature: "spat-trajstab"; Number of proposals: 71 tot: 1941
	Feature: "spat-trajstab-mbhstab"; Number of proposals: 66 tot: 2007
	Feature: "trajstab"; Number of proposals: 67 tot: 2074
	Feature: "trajstab-mbhstab"; Number of proposals: 73 tot: 2147

** Step 4: Compare clustered proposals of step 2 to ground-truth(s) to obtain Intersection over Union (IoU) scores per tube proposal **
python tubePy/runTube2iou.py example/clusts/001/nn10.hdf5 example/IoU/001/nn10T50.txt example/gt/001/tubes.hdf5 50 example/feat/001/idt.hdf5
	Read ground truth; 	reading tubes: example/gt/001/tubes.hdf5 ; Number of tubes to read: 1
	Get trajectory positions; 	Reading feature file: example/feat/001/idt.hdf5, "trajecShape" size (14337, 40)
	Get video dimensions; 	Reading feature file: example/feat/001/idt.hdf5, "vidinfo" size (3,)
	Feature: "hofstab", number of proposals: 48; 0 Best IoUs: [ 0.33105031] ; Best so far: [ 0.33105031]
	Feature: "hofstab-hogstab-trajstab-mbhstab", number of proposals: 70; 0 Best IoUs: [ 0.50275546] ; Best so far: [ 0.50275546]
	Feature: "hofstab-mbhstab", number of proposals: 62; 0 Best IoUs: [ 0.52908528] ; Best so far: [ 0.52908528]
	Feature: "hofstab-trajstab", number of proposals: 63; 0 Best IoUs: [ 0.49810889] ; Best so far: [ 0.52908528]
	Feature: "hofstab-trajstab-mbhstab", number of proposals: 67; 0 Best IoUs: [ 0.50223243] ; Best so far: [ 0.52908528]
	Feature: "hogstab", number of proposals: 71; 0 Best IoUs: [ 0.55509269] ; Best so far: [ 0.55509269]
	Feature: "hogstab-hofstab", number of proposals: 70; 0 Best IoUs: [ 0.53952658] ; Best so far: [ 0.55509269]
	Feature: "hogstab-hofstab-mbhstab", number of proposals: 69; 0 Best IoUs: [ 0.53791386] ; Best so far: [ 0.55509269]
	Feature: "hogstab-hofstab-trajstab", number of proposals: 66; 0 Best IoUs: [ 0.50420201] ; Best so far: [ 0.55509269]
	Feature: "hogstab-mbhstab", number of proposals: 71; 0 Best IoUs: [ 0.55121255] ; Best so far: [ 0.55509269]
	Feature: "hogstab-trajstab", number of proposals: 72; 0 Best IoUs: [ 0.50179219] ; Best so far: [ 0.55509269]
	Feature: "hogstab-trajstab-mbhstab", number of proposals: 70; 0 Best IoUs: [ 0.58327597] ; Best so far: [ 0.58327597]
	Feature: "mbhstab", number of proposals: 71; 0 Best IoUs: [ 0.53587353] ; Best so far: [ 0.58327597]
	Feature: "spat", number of proposals: 67; 0 Best IoUs: [ 0.54253024] ; Best so far: [ 0.58327597]
	Feature: "spat-hofstab", number of proposals: 69; 0 Best IoUs: [ 0.56753778] ; Best so far: [ 0.58327597]
	Feature: "spat-hofstab-mbhstab", number of proposals: 70; 0 Best IoUs: [ 0.56003076] ; Best so far: [ 0.58327597]
	Feature: "spat-hofstab-trajstab", number of proposals: 64; 0 Best IoUs: [ 0.55896378] ; Best so far: [ 0.58327597]
	Feature: "spat-hofstab-trajstab-mbhstab", number of proposals: 72; 0 Best IoUs: [ 0.55293119] ; Best so far: [ 0.58327597]
	Feature: "spat-hogstab", number of proposals: 70; 0 Best IoUs: [ 0.54906946] ; Best so far: [ 0.58327597]
	Feature: "spat-hogstab-hofstab", number of proposals: 75; 0 Best IoUs: [ 0.5381276] ; Best so far: [ 0.58327597]
	Feature: "spat-hogstab-hofstab-mbhstab", number of proposals: 75; 0 Best IoUs: [ 0.55357611] ; Best so far: [ 0.58327597]
	Feature: "spat-hogstab-hofstab-trajstab", number of proposals: 74; 0 Best IoUs: [ 0.54222506] ; Best so far: [ 0.58327597]
	Feature: "spat-hogstab-hofstab-trajstab-mbhstab", number of proposals: 72; 0 Best IoUs: [ 0.54427665] ; Best so far: [ 0.58327597]
	Feature: "spat-hogstab-mbhstab", number of proposals: 72; 0 Best IoUs: [ 0.544303] ; Best so far: [ 0.58327597]
	Feature: "spat-hogstab-trajstab", number of proposals: 74; 0 Best IoUs: [ 0.56068218] ; Best so far: [ 0.58327597]
	Feature: "spat-hogstab-trajstab-mbhstab", number of proposals: 76; 0 Best IoUs: [ 0.54952419] ; Best so far: [ 0.58327597]
	Feature: "spat-mbhstab", number of proposals: 70; 0 Best IoUs: [ 0.55278176] ; Best so far: [ 0.58327597]
	Feature: "spat-trajstab", number of proposals: 71; 0 Best IoUs: [ 0.55731547] ; Best so far: [ 0.58327597]
	Feature: "spat-trajstab-mbhstab", number of proposals: 66; 0 Best IoUs: [ 0.56513512] ; Best so far: [ 0.58327597]
	Feature: "trajstab", number of proposals: 67; 0 Best IoUs: [ 0.54874068] ; Best so far: [ 0.58327597]
	Feature: "trajstab-mbhstab", number of proposals: 73; 0 Best IoUs: [ 0.56846899] ; Best so far: [ 0.58327597]
	Writing IoU scores to example/IoU/001/nn10T50.txt

** Or do Step 4 on the exported tubes of step 3: Compare to ground-truth(s) to obtain Intersection over Union (IoU) scores per tube proposal **
python tubePy/runTube2iou.py example/tube/001/nn10T50.hdf5 example/IoU-tubes/001/nn10T50.txt example/gt/001/tubes.hdf5
	Read ground truth; 	reading tubes: example/gt/001/tubes.hdf5 ; Number of tubes to read: 1
	reading tubes: example/tube/001/nn10T50.hdf5 ; Number of tubes to read: 2147
	Feature: "tubes", number of proposals: 2147; 0 1000 2000 Best IoUs: [ 0.58327597] ; Best so far: [ 0.58327597]
	Writing IoU scores to example/IoU-tubes/001/nn10T50.txt

** Step 5: Find trajectory IDs that are inside of each clustered proposal from step 2 **
python tubePy/runTube2trajIDs.py example/clusts/001/nn10.hdf5 example/trajIDs/001/nn10T50.hdf5 example/feat/001/idt.hdf5 50
	Get trajectory positions; 	Reading feature file: example/feat/001/idt.hdf5, "trajecShape" size (14337, 40)
	Get video dimensions; 	Reading feature file: example/feat/001/idt.hdf5, "vidinfo" size (3,)
	Feature: "hofstab", number of proposals: 48;
	Feature: "hofstab-hogstab-trajstab-mbhstab", number of proposals: 70;
	Feature: "hofstab-mbhstab", number of proposals: 62;
	Feature: "hofstab-trajstab", number of proposals: 63;
	Feature: "hofstab-trajstab-mbhstab", number of proposals: 67;
	Feature: "hogstab", number of proposals: 71;
	Feature: "hogstab-hofstab", number of proposals: 70;
	Feature: "hogstab-hofstab-mbhstab", number of proposals: 69;
	Feature: "hogstab-hofstab-trajstab", number of proposals: 66;
	Feature: "hogstab-mbhstab", number of proposals: 71;
	Feature: "hogstab-trajstab", number of proposals: 72;
	Feature: "hogstab-trajstab-mbhstab", number of proposals: 70;
	Feature: "mbhstab", number of proposals: 71;
	Feature: "spat", number of proposals: 67;
	Feature: "spat-hofstab", number of proposals: 69;
	Feature: "spat-hofstab-mbhstab", number of proposals: 70;
	Feature: "spat-hofstab-trajstab", number of proposals: 64;
	Feature: "spat-hofstab-trajstab-mbhstab", number of proposals: 72;
	Feature: "spat-hogstab", number of proposals: 70;
	Feature: "spat-hogstab-hofstab", number of proposals: 75;
	Feature: "spat-hogstab-hofstab-mbhstab", number of proposals: 75;
	Feature: "spat-hogstab-hofstab-trajstab", number of proposals: 74;
	Feature: "spat-hogstab-hofstab-trajstab-mbhstab", number of proposals: 72;
	Feature: "spat-hogstab-mbhstab", number of proposals: 72;
	Feature: "spat-hogstab-trajstab", number of proposals: 74;
	Feature: "spat-hogstab-trajstab-mbhstab", number of proposals: 76;
	Feature: "spat-mbhstab", number of proposals: 70;
	Feature: "spat-trajstab", number of proposals: 71;
	Feature: "spat-trajstab-mbhstab", number of proposals: 66;
	Feature: "trajstab", number of proposals: 67;
	Feature: "trajstab-mbhstab", number of proposals: 73;
	Write trajectory IDs to example/trajIDs/001/nn10T50.hdf5

** Or do Step 5 on the exported tubes of step 3: Find trajectory IDs that are inside of each proposal **
python tubePy/runTube2trajIDs.py example/tube/001/nn10T50.hdf5 example/trajIDs-tubes/001/nn10T50.hdf5 example/feat/001/idt.hdf5
	Get trajectory positions; 	Reading feature file: example/feat/001/idt.hdf5, "trajecShape" size (14337, 40)
	reading tubes: example/tube/001/nn10T50.hdf5 ; Number of tubes to read: 2147
	Feature: "tubes", number of proposals: 2147;
	Write trajectory IDs to example/trajIDs-tubes/001/nn10T50.hdf5

isTrimmedVideo=1;

鲜花

握手

雷人

路过

鸡蛋

该文章已有0人参与评论

请发表评论

全部评论

专题导读

More+

10-27 六六分期app的软件客服如何联系？(六六分期

11-06 可心卡盟:win10系统火狐flash插件崩溃怎么

11-06 亲亲特价:怎么删除回收站图标

11-06 济南大学虚拟社区:鲁大师节能降温的具体办

11-06 xlueops.exe:无线网络安装向导

11-06 女斗合众国:win7系统cf与主机连接不稳定怎

11-06 0xc000022-[cf烟雾头]cf怎么调烟雾头

11-06 qizideyouhuo:应用程序无法正常启动0xc0000

11-06 ipz-185:win7系统vcf文件怎么打开

11-06 傻哥蹦迪:win10系统s4怎么打开usb调试

11-06 八神浩树gtaste:回收站清空了怎么恢复

11-06 妖尾之黑色守护:win10系统电脑没有1440x900

11-06 校园至尊魔王小说:win7系统浏览网页时字体

11-06 女斗合众国:win10系统访问共享文件夹提示请

11-06 tokyo hot n0654:恢复win7系统默认字体一招

11-06 雨酷仙境:设置win7系统转移临时文件夹腾出

11-06 阿穆纳伊之杖:win7系统开始菜单在右边还原

11-06 tunespotting:win10系统火狐flash插件总是

11-06 甘尔葛分析师：计谋网站seo关键词暴涨有什

11-06 蔡贵霖: 计谋网站seo关键词暴涨有什么秘密

11-06 博益网首页:ao3网页版进入不了解决方法

11-06 漏斗子专栏: 网站数据分析小白易懂精华篇

11-06 见证双虹怎么做:win7系统开启telnet命令的

11-06 颾狐蝶蜋:系统资源不足无法完成请求的服务

11-06 国光中学校歌:提交网站到alexa查询详细步骤

11-06 西安有情天:静态网页和动态网页的区别

11-06 红木雅尚斋:外部链接构造对网站的好处

11-06 前官礼遇：防止域名劫持–增强域安全性的10

11-06 密传二转答案: 中文分词算法有哪些

11-06 金泉家园邮编:百度快照劫持的表现及应对方

naezith/naezith-localization: Community project to have many languages in Remnan ...发布时间：2022-08-16

xwl/cal-china-x: chinese localization, prioritized holiday highlight, lunar, hor ...发布时间：2022-08-16

剪的笔顺,诠释剪的笔画,认识剪的部首

1 六六分期app的软件客服如何联系？(六六分期

六六分期app的软件客服如何联系？不知道吗？加qq群【895510560】即可！标题：六六分期

阅读：19230|2023-10-27

2 可心卡盟:win10系统火狐flash插件崩溃怎么

今天小编告诉大家如何处理win10系统火狐flash插件总是崩溃的问题，可能很多用户都不知

阅读：9996|2022-11-06

3 亲亲特价:怎么删除回收站图标

今天小编告诉大家如何对win10系统删除桌面回收站图标进行设置，可能很多用户都不知道

阅读：8331|2022-11-06

4 济南大学虚拟社区:鲁大师节能降温的具体办

今天小编告诉大家如何对win10系统电脑设置节能降温的设置方法，想必大家都遇到过需要

阅读：8700|2022-11-06

5 xlueops.exe:无线网络安装向导

我们在使用xp系统的过程中,经常需要对xp系统无线网络安装向导设置进行设置，可能很多

阅读：8645|2022-11-06

6 女斗合众国:win7系统cf与主机连接不稳定怎

今天小编告诉大家如何处理win7系统玩cf老是与主机连接不稳定的问题，可能很多用户都不

阅读：9667|2022-11-06

7 0xc000022-[cf烟雾头]cf怎么调烟雾头

电脑对日常生活的重要性小编就不多说了，可是一旦碰到win7系统设置cf烟雾头的问题，很

阅读：8630|2022-11-06

8 qizideyouhuo:应用程序无法正常启动0xc0000

我们在日常使用电脑的时候，有的小伙伴们可能在打开应用的时候会遇见提示应用程序无法

阅读：8004|2022-11-06

9 ipz-185:win7系统vcf文件怎么打开

今天小编告诉大家如何对win7系统打开vcf文件进行设置，可能很多用户都不知道怎么对win

阅读：8664|2022-11-06

10 傻哥蹦迪:win10系统s4怎么打开usb调试

今天小编告诉大家如何对win10系统s4开启USB调试模式进行设置，可能很多用户都不知道怎

阅读：7540|2022-11-06

扫描微信二维码

查看手机版网站

随时了解更新最新资讯

139-2527-9053

在线客服（服务时间 9:00～18:00）

在线QQ客服

地址：深圳市南山区西丽大学城创智工业园

电邮：jeky_zhao#qq.com

移动电话：139-2527-9053

客服电话

电子邮件

jvgemert/apt: Action localization Proposals from dense Trajectories

开源软件名称（OpenSource Name）：

开源软件地址(OpenSource Url)：

开源编程语言(OpenSource Language)：

开源软件介绍(OpenSource Introduction)：

APT: Action localization Proposals from dense Trajectories

Requirements:

Cython:

Organization

doExample.sh

Input and output

Frequently Asked Questions

- Did the UCF101 scores change?

- Do you have pre-computed tubes available?

UCF-Sports (trimmed)

UCF-101 (trimmed)

MSR Action Dataset II (untrimmed)

- Do you also have the 'Tubelet' results?

- What is the tube dataformat?

- How do I read a tube from disk?

Matlab:

Python, using the provide TubeList class:

Python, without the TubeList class:

- What does the output look like?

isTrimmedVideo=0;

isTrimmedVideo=1;

请发表评论

全部评论

上一篇：

下一篇：

CVE-2022-23003

krishnaik06/Machine-Learning-in-90-days

armancodv/building-energy-model-matlab:

美元符号为什么是“$”

FGRibreau/import-tweets-to-mastodon: How

剪的笔顺,诠释剪的笔画,认识剪的部首

六六分期app的软件客服如何联系？(六六分期

florent37/ViewAnimator: A fluent Android

florent37/Shrine-MaterialDesign2: implem

CVE-2020-36276

SimpleSoftwareIO/simple-sms: Send and re

关于我们

产品与服务

解决方案

139-2527-9053