shaoshengsong / DeepSORT

support deepsort and bytetrack MOT(Multi-object tracking) using yolov5 with C++
777 stars 182 forks source link

代码运行到 FeatureTensor.cpp 第101行不动了 #8

Open z19940829i opened 4 years ago

z19940829i commented 4 years ago

您好,您的代码运行到 seesion->Run(feed_dict,outnames,{},&output_tensors),时,为什么一直阻塞在这里不动了,另外模型文件我从网上下载的是mars-small128.ckpt-68577.meta,和您的mars-small128.meta是不是一样的?

Scheaven commented 4 years ago

您解决了吗?我也到这一步卡住了,我的环境是跟作者的不一样的。我用的tf 1.13.1

z19940829i commented 4 years ago

还没解决,我换了个模型,好像能跑起来了,但是结果不对发自我的华为手机-------- 原始邮件 --------发件人: Schro Heaven notifications@github.com日期: 2020年4月1日周三 19:39收件人: shaoshengsong/DeepSORT DeepSORT@noreply.github.com抄送: z19940829i zhangyuwei001@126.com, Author author@noreply.github.com主 题: Re: [shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)

您解决了吗?我也到这一步卡住了,我的环境是跟作者的不一样的。我用的tf 1.13.1

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe.

z19940829i commented 4 years ago

我用的tensorflow1.8,模型在网上随便找了个,不是用从deepsort官网上那里下的,然后跑到那里至少不会阻塞了发自我的华为手机-------- 原始邮件 --------发件人: zhangyuwei001@126.com日期: 2020年4月1日周三 20:49收件人: shaoshengsong/DeepSORT reply@reply.github.com主 题: 回复:[shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)还没解决,我换了个模型,好像能跑起来了,但是结果不对发自我的华为手机-------- 原始邮件 --------发件人: Schro Heaven notifications@github.com日期: 2020年4月1日周三 19:39收件人: shaoshengsong/DeepSORT DeepSORT@noreply.github.com抄送: z19940829i zhangyuwei001@126.com, Author author@noreply.github.com主 题: Re: [shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)

您解决了吗?我也到这一步卡住了,我的环境是跟作者的不一样的。我用的tf 1.13.1

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe.

z19940829i commented 4 years ago

我想是因为我用在了人脸跟踪那里,而不是行人跟踪,会不会就提取不了特征了?发自我的华为手机-------- 原始邮件 --------发件人: zhangyuwei001@126.com日期: 2020年4月1日周三 20:49收件人: shaoshengsong/DeepSORT reply@reply.github.com主 题: 回复:[shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)还没解决,我换了个模型,好像能跑起来了,但是结果不对发自我的华为手机-------- 原始邮件 --------发件人: Schro Heaven notifications@github.com日期: 2020年4月1日周三 19:39收件人: shaoshengsong/DeepSORT DeepSORT@noreply.github.com抄送: z19940829i zhangyuwei001@126.com, Author author@noreply.github.com主 题: Re: [shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)

您解决了吗?我也到这一步卡住了,我的环境是跟作者的不一样的。我用的tf 1.13.1

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe.

Scheaven commented 4 years ago

我想是因为我用在了人脸跟踪那里,而不是行人跟踪,会不会就提取不了特征了?发自我的华为手机-------- 原始邮件 --------发件人: zhangyuwei001@126.com日期: 2020年4月1日周三 20:49收件人: shaoshengsong/DeepSORT reply@reply.github.com主 题: 回复:[shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)还没解决,我换了个模型,好像能跑起来了,但是结果不对发自我的华为手机-------- 原始邮件 --------发件人: Schro Heaven notifications@github.com日期: 2020年4月1日周三 19:39收件人: shaoshengsong/DeepSORT DeepSORT@noreply.github.com抄送: z19940829i zhangyuwei001@126.com, Author author@noreply.github.com主 题: Re: [shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8) 您解决了吗?我也到这一步卡住了,我的环境是跟作者的不一样的。我用的tf 1.13.1 —You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe.

应该不会,特征应该都能够提取。 你的意思是说代码是对的,只是训练好的模型没办法运行时吗? 能给发一下您找的模型或者链接吗?邮箱 506321815@qq.com 微信也可以(号: 506321815)

Scheaven commented 4 years ago

您好:      能把你找到的跟踪度量模型分享一下吗?      我这边着急用,谢谢!      有机会希望可以和您讨论问题!!

------------------ 原始邮件 ------------------ 发件人: "z19940829i"<notifications@github.com>; 发送时间: 2020年4月1日(星期三) 晚上8:53 收件人: "shaoshengsong/DeepSORT"<DeepSORT@noreply.github.com>; 抄送: "Scheaven"<506321815@qq.com>;"Comment"<comment@noreply.github.com>; 主题: Re: [shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)

我想是因为我用在了人脸跟踪那里,而不是行人跟踪,会不会就提取不了特征了?发自我的华为手机-------- 原始邮件 --------发件人: zhangyuwei001@126.com日期: 2020年4月1日周三 20:49收件人: shaoshengsong/DeepSORT <reply@reply.github.com>主 题: 回复:[shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)还没解决,我换了个模型,好像能跑起来了,但是结果不对发自我的华为手机-------- 原始邮件 --------发件人: Schro Heaven <notifications@github.com>日期: 2020年4月1日周三 19:39收件人: shaoshengsong/DeepSORT <DeepSORT@noreply.github.com>抄送: z19940829i <zhangyuwei001@126.com>, Author <author@noreply.github.com>主 题: Re: [shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8) 您解决了吗?我也到这一步卡住了,我的环境是跟作者的不一样的。我用的tf 1.13.1

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

qiaoshh commented 4 years ago

您好:      能把你找到的跟踪度量模型分享一下吗?      我这边着急用,谢谢!      有机会希望可以和您讨论问题!! ------------------ 原始邮件 ------------------ 发件人: "z19940829i"<notifications@github.com>; 发送时间: 2020年4月1日(星期三) 晚上8:53 收件人: "shaoshengsong/DeepSORT"<DeepSORT@noreply.github.com>; 抄送: "Scheaven"<506321815@qq.com>;"Comment"<comment@noreply.github.com>; 主题: Re: [shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8) 我想是因为我用在了人脸跟踪那里,而不是行人跟踪,会不会就提取不了特征了?发自我的华为手机-------- 原始邮件 --------发件人: zhangyuwei001@126.com日期: 2020年4月1日周三 20:49收件人: shaoshengsong/DeepSORT <reply@reply.github.com>主 题: 回复:[shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8)还没解决,我换了个模型,好像能跑起来了,但是结果不对发自我的华为手机-------- 原始邮件 --------发件人: Schro Heaven <notifications@github.com>日期: 2020年4月1日周三 19:39收件人: shaoshengsong/DeepSORT <DeepSORT@noreply.github.com>抄送: z19940829i <zhangyuwei001@126.com>, Author <author@noreply.github.com>主 题: Re: [shaoshengsong/DeepSORT] 代码运行到 FeatureTensor.cpp 第101行不动了 (#8) 您解决了吗?我也到这一步卡住了,我的环境是跟作者的不一样的。我用的tf 1.13.1 —You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. This repository runs fine, if you want, i can share it with your.

doubleLLL3 commented 3 years ago

前期

本来和上面说的一样,在FeatureTensor.cpp文件的下面这个位置卡住不动了: image 后来我在这个网址下载了mars-small128.ckpt-68577.meta这个模型,可以顺利运行。

问题

不停地报Tensorflow get feature succeed!警告。 image 查看status,信息为节点不匹配 image 节点设置为作者默认: image

自己的尝试

在init函数里的下面打印结点的代码,在我这里无法运行,要是能打印出所有的结点也许会有帮助。

for (const auto &node : graph_def.graph_def().node()) {
        printf("node name:%s\n", node.name().c_str());
    }

PS:因为在我的tensorflow1.10环境里,原先的代码编译会报错 image 所以根据这篇博客tensorflow1.8 c++ 加载训练好的模型踩坑-CSDN修改了部分代码: image 主要是修改init函数中,graph_def()→mutable_graph_def(),saver_def()→mutable_saver_def(),注意指针。

但在打印节点这部分不知道如何修改,上面的博客里没有提到。

期待大佬们的解决~

FannyDuan commented 3 years ago

您解决了吗?我也到这一步卡住了,我的环境是跟作者的不一样的。我用的tf 1.13.1

您好,请问您后来解决这个问题了吗?