【openCV】使用opencv_traincascade.exe 遇到的问题

2013-03-11 01:50:01 6.9k 阅读 AI 写同款 · GEO 优化AI 选题数据监测一键分发 ›

采用OpenCv的traincascade.exe来训练自己的分类器，用于目标的检测 opencv中有两个函数可以训练分类器opencv_haartraining.exe和opencv_traincascade.exe，前者只能训练haar特征，后者可以用HAAR、LBP和HOG特征训练分类器。这两个函数都可以在opencv\build\x86\vc10\bin文件夹下找到。opencv_traincascade.exe训练的是adaboost的级联分类器，这里不讲adaboost... 阅读详情

各位高人

想用opencv_traincascade.exe 训练自己的XML，简要步骤如下

1， createsamples
其中使用的posNum 5709, negNum 4381，其中所有的pos, neg的大小都是 20*20
训练出来的vec data大小有 4464 KB

2, 第一次使用 opencv_traincascade.exe -data D:\traincascade -vec D:\data\pos.vec -bg D:\bg.txt -numPos 5709-numNeg 4381 -numStages 22 -precalcValBufSize 200 -precalcIdxBufSize 1000 -featureType LBP -w 20 -h 20，会有有一个断言失败，opencv_traincascade.exe OpenCV Error: Assertion failed (elements_read == 1)

之后我将相应的numPos 调小至 5000， numNeg 调至4350，断言消失，很快就会产生7 个parenet Node, 但在第8个stage, 一直就停住了，跑了2天还是如此

不知道具体的原因是什么？求解？多谢各位
或者说createsamples中的numpos, 与negNum 的值与traincascade 中的对应的值有什么关系？比例是怎么样的

...全文

opencv_traincascade问题在用opencv_traincascade训练分类器的时候，遇到了报错如下： Train dataset for temp stage can not be filled. Branch training terminated. 在stackoverflow上查了一下，问题的根源在于负样本读取失败，导致问题的原因有： 1）负样本描述文件neg.txt不能带路径名，即： -bg 阅读详情

iwill_chyw

机器视觉 4,514 成员 15,347 内容

+ 关注 + 加入社区

切换为时间正序

LHY_045 2015-03-02

打赏
举报

楼主您问题解决了吗？我现在一输入opencv_traincascade -data dt -vec train.vec -bg bg.txt -numPos 1500 -numNeg 3000 -w 120 -h 120的语句就报错，不知为何。

hangzadfg 2014-12-08

打赏
举报

-numPos参数指定的是在一个stage中使用的正样本个数，并不是全部正样本个数

上岸の鱼 2014-11-04

打赏
举报

引用 22 楼 a6670369 的回复:

[quote=引用 6 楼 cjnaning 的回复:] 这是我在网上搜到的正解： The problem is that your vec-file has exactly the same samples count that you passed in command line -numPos 979. Training application used all samples from the vec-file to train 0-stage and it can not get new positive samples for the next stage training because vec-file is over. The bug of traincascade is that it had assert() in such cases, but it has to throw an exception with error message for a user. It was fixed in r8913. -numPose is a samples count that is used to train each stage. Some already used samples can be filtered by each previous stage (ie recognized as background), but no more than (1 - minHitRate) * numPose on each stage. So vec-file has to contain >= (numPose + (numStages-1) * (1 - minHitRate) * numPose) + S, where S is a count of samples from vec-file that can be recognized as background right away. I hope it can help you to create vec-file of correct size and chose right numPos value. It worked for me. I also had same problem, I was following the famous tutorial on HAAR training but wanted to try the newer training utility with -npos 7000 -nneg 2973 so i did following calcs: vec-file has to contain >= (numPos + (numStages-1) * (1 - minHitRate) * numPos) + S 7000 >= (numPos + (20-1) * (1 - 0.999) * numPos) + 2973 (7000 - 2973)/(1 + 19*0.001) >= numPos numPos <= 4027/1.019 numPos <= 3951 ~~ 3950 and used: -npos 3950 -nneg 2973 It works.

这些话是什么意思啊，我看不懂，，，我总是有这个错误Traincascade Error: Bad argument (Can not get new positive sample. The most possible reason is insufficient count of samples in given vec-file.，知道怎么解决么[/quote] 这句话是说：最可能的原因是你的样本数量不够。

小ks强 2014-08-29

打赏
举报

引用 6 楼 cjnaning 的回复:

这是我在网上搜到的正解： The problem is that your vec-file has exactly the same samples count that you passed in command line -numPos 979. Training application used all samples from the vec-file to train 0-stage and it can not get new positive samples for the next stage training because vec-file is over. The bug of traincascade is that it had assert() in such cases, but it has to throw an exception with error message for a user. It was fixed in r8913. -numPose is a samples count that is used to train each stage. Some already used samples can be filtered by each previous stage (ie recognized as background), but no more than (1 - minHitRate) * numPose on each stage. So vec-file has to contain >= (numPose + (numStages-1) * (1 - minHitRate) * numPose) + S, where S is a count of samples from vec-file that can be recognized as background right away. I hope it can help you to create vec-file of correct size and chose right numPos value. It worked for me. I also had same problem, I was following the famous tutorial on HAAR training but wanted to try the newer training utility with -npos 7000 -nneg 2973 so i did following calcs: vec-file has to contain >= (numPos + (numStages-1) * (1 - minHitRate) * numPos) + S 7000 >= (numPos + (20-1) * (1 - 0.999) * numPos) + 2973 (7000 - 2973)/(1 + 19*0.001) >= numPos numPos <= 4027/1.019 numPos <= 3951 ~~ 3950 and used: -npos 3950 -nneg 2973 It works.

小ks强 2014-08-29

打赏
举报

引用 20 楼 woaipaoche 的回复:

同志，有没有opencv_traincascade.exe给我一份，发到我邮箱xinxueab@163.com，我的是opencv2.4.6版本怎么安装后找不到呢？跑到哪里去了呢？？、、、

我给你发了，记得接收

天使之一 2014-08-14

打赏
举报

同志，有没有opencv_traincascade.exe给我一份，发到我邮箱xinxueab@163.com，我的是opencv2.4.6版本怎么安装后找不到呢？跑到哪里去了呢？？、、、

思飞工作室 2014-08-13

打赏
举报

引用 12 楼 a201c501ys 的回复:

opencv_haartraining.exe 训练某一层停住时需要更新负样本，按ctrl+c 输入y停止，换新的相同数量的负样本就行了，已经训练成功几次了。

具体怎么个换法？？

blueink_200451 2014-06-29

打赏
举报

yundanfengqing1qq 2014-06-27

打赏
举报

这个当时我在后面加了 —stages 4，我在第6层卡住了，因此只让前5层生成的强分类器（0,1,2,3,4）生成XML文件，但是对于那种不卡住的，系统自动生成xml的，我也没有弄出来

yundanfengqing1qq 2014-06-27

打赏
举报

既然没有认错的样本了，那为什么卡主，全部分类正确后，应该正常退出啊

宁静致远_淡泊明志 2014-05-01

打赏
举报

我也遇到了同样的问题，一星期了，有解决的吗？非常感谢

蓝色的杯子 2014-04-14

打赏
举报

楼主没有加源码调么，这些小问题你一调就知道了。停止没反应是木有认错的负本了

蓝色的杯子 2014-04-08

打赏
举报

断言是样本太少了，我们的没有并行，要训练3天

a201c501ys 2014-04-06

打赏
举报

opencv_haartraining.exe 训练某一层停住时需要更新负样本，按ctrl+c 输入y停止，换新的相同数量的负样本就行了，已经训练成功几次了。

mdq_123 2014-03-13

打赏
举报

楼主，负样本是怎么做呢？直接在网上随便找些图片？

meng1991zhong 2014-02-27

打赏
举报

楼主不好意思打扰了为什么我找不到opencv_traincascade.exe呢？我用的是OPENCV2.4.3

andyboliu 2013-09-02

打赏
举报

读取或处理到了没有数据的内存，看一下读取或者处理的图像大小，数目设置哪里超出了实际值。

天使之一 2013-08-20

打赏
举报

很遗憾的说我也遇到这个问题，出现了断言失败，错误地点在opencv/modules/core/src/arithm。cpp中 2841行

heathyhuhu 2013-08-14

打赏
举报

为什么我的总是报那个nullpointer 的 Null pointer (NULL filename) in unknown function, file ..\..\..\src persistence 2603 错误，这个命令下的参数是严格按照参考文件来的，为什么会报错？？？在线急等啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊啊！

cjnaning 2013-05-07

打赏
举报

这是我在网上搜到的正解： The problem is that your vec-file has exactly the same samples count that you passed in command line -numPos 979. Training application used all samples from the vec-file to train 0-stage and it can not get new positive samples for the next stage training because vec-file is over. The bug of traincascade is that it had assert() in such cases, but it has to throw an exception with error message for a user. It was fixed in r8913. -numPose is a samples count that is used to train each stage. Some already used samples can be filtered by each previous stage (ie recognized as background), but no more than (1 - minHitRate) * numPose on each stage. So vec-file has to contain >= (numPose + (numStages-1) * (1 - minHitRate) * numPose) + S, where S is a count of samples from vec-file that can be recognized as background right away. I hope it can help you to create vec-file of correct size and chose right numPos value. It worked for me. I also had same problem, I was following the famous tutorial on HAAR training but wanted to try the newer training utility with -npos 7000 -nneg 2973 so i did following calcs: vec-file has to contain >= (numPos + (numStages-1) * (1 - minHitRate) * numPos) + S 7000 >= (numPos + (20-1) * (1 - 0.999) * numPos) + 2973 (7000 - 2973)/(1 + 19*0.001) >= numPos numPos <= 4027/1.019 numPos <= 3951 ~~ 3950 and used: -npos 3950 -nneg 2973 It works.

加载更多回复（5）

【openCV】使用opencv_traincascade.exe 遇到的问题

相关推荐