<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[RZJMPB的博客]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/RZJMPB</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; RZJMPB]]></copyright><item><title><![CDATA[pytorch 训练正常,测试显存爆掉问题解决]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/85684123</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/85684123</guid><author>RZJMPB</author><pubDate>Thu, 03 Jan 2019 11:27:43 +0800</pubDate><description><![CDATA[pytorch 训练正常,测试显存爆掉问题解决Intropytorch versioncudnn versionpytorch训练正常但是在预测的时候显存爆掉,错误信息如下暂时解决方法:[参考链接](https://discuss.pytorch.org/t/strange-gpu-memory-behavior-strange-memory-consumption-and-out-of-memo...]]></description><category></category></item><item><title><![CDATA[MaxPooling的作用 and some tips about CNN]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/52550582</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/52550582</guid><author>RZJMPB</author><pubDate>Thu, 15 Sep 2016 19:03:30 +0800</pubDate><description><![CDATA[摘抄记录： 
MaxPooling 
Another important concept of CNNs is max-pooling, which is a form of non-linear down-sampling. Max-pooling partitions the input image into a set of non-overlapping rectangles and, fo]]></description><category></category></item><item><title><![CDATA[解决Jetson Tegra TK1 编译 YOLO Darknet出现的不支持，compute_52 错误]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/52448344</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/52448344</guid><author>RZJMPB</author><pubDate>Tue, 06 Sep 2016 11:24:55 +0800</pubDate><description><![CDATA[修改Makefile: 
源代码： 
ARCH= –gpu-architecture=compute_52 –gpu-code=compute_52改为： 
ARCH= –gpu-architecture=compute_20 –gpu-code=compute_20]]></description><category></category></item><item><title><![CDATA[训练Faster R-CNN出现的errors]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/52415030</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/52415030</guid><author>RZJMPB</author><pubDate>Fri, 02 Sep 2016 16:30:20 +0800</pubDate><description><![CDATA[1.  assert (boxes[:, 2] >= boxes[:, 0]).all() AssertionError目前临时解决的方法： 
在 imdb.py中：for b in range(len(boxes)):     
                if boxes[b][2] < boxes[b][0]:
                   boxes[b][0] = 0asser]]></description><category></category></item><item><title><![CDATA[在Faster R-CNN 中DEMO 的CPU（i5）和GPU（GTX1060 ）时间对比]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/52382403</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/52382403</guid><author>RZJMPB</author><pubDate>Wed, 31 Aug 2016 10:27:04 +0800</pubDate><description><![CDATA[软硬件环境：Ubuntu 14.04 64bit  
NVIDIA GeForce GTX 1060 6GB 
Intel® Core™ i5-6500 CPU @ 3.20GHz × 4  
8GB memory 
CUDA 8.0  
cuDNN 5.0 zf net使用CPU的时间：~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Demo for data/demo/0]]></description><category></category></item><item><title><![CDATA[Faster R-CNN 和最新的版本cuDNN V5.0不兼容问题]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/52373012</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/52373012</guid><author>RZJMPB</author><pubDate>Tue, 30 Aug 2016 21:09:12 +0800</pubDate><description><![CDATA[Faster R-CNN 和最新的版本cuDNN V5.0不兼容问题软件硬件环境: 
 Ubuntu 14.04 64bit 
 NVIDIA GTX1060 
cuda 8.0 
cuDNN 5.0 
错误如下：解决方法：cd py-faser-rcnn/caffe-fast-rcnn  
Git remote add caffe https://github.com/BVLC/caffe.git]]></description><category></category></item><item><title><![CDATA[二维数组的动态内存分配]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50984460</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50984460</guid><author>RZJMPB</author><pubDate>Fri, 25 Mar 2016 22:37:15 +0800</pubDate><description><![CDATA[#include <malloc.h>
#include <iostream>using namespace std;int main(int argc, char **argv)
{
    int n;
    int m=2;
    cout << "请输入通信录的个数： " << endl;    cin >> n;    int **memo;
    memo = (int**) ma]]></description><category></category></item><item><title><![CDATA[可否强制修改常量的值？]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50982220</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50982220</guid><author>RZJMPB</author><pubDate>Fri, 25 Mar 2016 16:53:36 +0800</pubDate><description><![CDATA[#include <iostream>
using namespace std;int main()
{
    int *p;    const int DATA = 100;    p = (int*) &DATA;   //(int*) 此处需要强制类型转换 因为&DATA的类型为const int* 
    cout << "Old *P is " << *p << endl;
    *]]></description><category></category></item><item><title><![CDATA[C++分离类模板的声明和定义]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50929868</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50929868</guid><author>RZJMPB</author><pubDate>Sat, 19 Mar 2016 10:43:22 +0800</pubDate><description><![CDATA[成员函数在体外的定义格式为： 
template  返回值类型 类名::成员函数名（参数列表） 
{ 
//类模板内容 
}；实例：#include <iostream>using namespace std;template <class T> class MAX    //类模板 
{
    public:
        MAX(T v1, T v2);
        T getmax(M]]></description><category></category></item><item><title><![CDATA[并行编程1：CUDA基础]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50907780</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50907780</guid><author>RZJMPB</author><pubDate>Wed, 16 Mar 2016 19:40:36 +0800</pubDate><description><![CDATA[CUDA（Compute Unified Device Architecture，统一计算架构[1]）是由NVIDIA所推出的一种集成技术，是该公司对于GPGPU的正式名称。通过这个技术，用户可利用NVIDIA的GeForce 8以后的GPU和较新的Quadro GPU进行计算。亦是首次可以利用GPU作为C-编译器的开发环境。NVIDIA营销的时候[2]，往往将编译器与架构混合推广，造成混乱。实际上]]></description><category></category></item><item><title><![CDATA[拷贝构造函数]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50897653</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50897653</guid><author>RZJMPB</author><pubDate>Tue, 15 Mar 2016 16:33:08 +0800</pubDate><description><![CDATA[A *foo(A a)
{
    cout << "-----foo enter-----"<< endl;
    A *p = new A();
    cout << "-----foo exit------"<< endl;
    return p;
}A *boo(const A& a) 
{ 
     cout << “———-boo enter——–”<]]></description><category></category></item><item><title><![CDATA[【ZYNQ-7000开发之十六】音频信号处理]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50897563</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50897563</guid><author>RZJMPB</author><pubDate>Tue, 15 Mar 2016 16:25:07 +0800</pubDate><description><![CDATA[本编文章将使用Zynq开发平台Miz702上的ADAU1761音频编解码芯片，进行采样模拟音频信号（2路），转成数字信号，然后再通过ADAU1761转成模拟信号输出（2路）。在此基础上通过左右声道相减，简单的实现人声消除。本文不介绍ADAU1761驱动的编写方法，直接采用 Ali Aljaani编写的IP。 
ADAU1761参数的配置参考 ADA1761 Datasheet本文所使用的开发板是Mi]]></description><category></category></item><item><title><![CDATA[【ZYNQ-7000开发之十五】Vivado HLS和OpenCV_下]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50858149</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50858149</guid><author>RZJMPB</author><pubDate>Fri, 11 Mar 2016 18:40:41 +0800</pubDate><description><![CDATA[本篇文章在上篇文章【ZYNQ-7000开发之十四】的基础上，利用OpenCV HLS的函数，hls::Sobel,编写边缘检测算法，仿真测试，创建IP,最后在vivado里测试Sobel IP，此外本篇文章的实验还要基于 【ZYNQ-7000开发之十】TGP+VDMA+HDMI搭建视频通路]]></description><category></category></item><item><title><![CDATA[【ZYNQ-7000开发之十四】Vivado HLS和OpenCV_上]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50829932</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50829932</guid><author>RZJMPB</author><pubDate>Tue, 08 Mar 2016 19:46:06 +0800</pubDate><description><![CDATA[本篇文章对Vivado HLS以及OpenCV HLS进行简要介绍，并且利用OpenCV HLS的函数，hls::Sobel,编写边缘检测算法，仿真测试，创建IP,最后在vivado里测试Sobel IP关于HLS详细内容请参考官方文献ug902-vivado-high-level-synthesis 
*本文主要参考文献： 
how_to_accelerate_opencv_application]]></description><category></category></item><item><title><![CDATA[【ZYNQ-7000开发之十三】中断：私有定时器中断]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50812579</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50812579</guid><author>RZJMPB</author><pubDate>Sun, 06 Mar 2016 11:32:01 +0800</pubDate><description><![CDATA[上篇文章实现了了PS接受来自PL的中断，本片文章将在ZYNQ的纯PS里实现私有定时器中断。每个一秒中断一次，在中断函数里计数加1，通过串口打印输出。*本文所使用的开发板是Miz702(兼容zedboard)  
PC 开发环境版本：Vivado 2015.2 Xilinx SDK 2015.2*中断原理中断对于保证任务的实时性非常必要，在ZYNQ里集成了中断控制器GIC(Generic Interr]]></description><category></category></item><item><title><![CDATA[【ZYNQ-7000开发之十二】中断：PS接收来自PL的中断]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50736918</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50736918</guid><author>RZJMPB</author><pubDate>Thu, 25 Feb 2016 13:10:32 +0800</pubDate><description><![CDATA[本篇文章主要介绍外设（PL）产生的中断请求，在PS端进行处理。 
在PL端通过按键产生中断，PS接受到之后点亮相应的LED.本文所使用的开发板是Miz702(兼容zedboard)  
PC 开发环境版本：Vivado 2015.2 Xilinx SDK 2015.2 搭建硬件工程建好工程后，添加ZYNQ IP双击 ZYNQ,打开Re-customize IP对话框，使能IRQ_P2P使能UART1]]></description><category></category></item><item><title><![CDATA[【ZYNQ-7000开发之十一】VGA RLT代码封装成AXI Stream]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50609314</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50609314</guid><author>RZJMPB</author><pubDate>Fri, 29 Jan 2016 21:28:05 +0800</pubDate><description><![CDATA[本编文章将对VGA的RTL代码，封装成AXI Stream,并且在vivado 里用TPG进行测试本篇文章的VGA RTL代码在【ZYNQ-7000开发之一】基础上修改，封装好的VGA Stream可以方面我们实现视频图像处理]]></description><category></category></item><item><title><![CDATA[【ZYNQ-7000开发之十】TGP+VDMA+HDMI搭建视频通路]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50589411</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50589411</guid><author>RZJMPB</author><pubDate>Tue, 26 Jan 2016 19:32:56 +0800</pubDate><description><![CDATA[本编文章在ADI官方HDMI例程的基础上进行修改，实现视频通路，为使用ZYNQ视频处理做好必要准备。在 【ZYNQ-7000开发之九】使用VDMA在PL和PS之间传输视频流数据 这篇文章中，介绍了如何使用VDMA传输stream类型的视频流数据，本次实验将结合【ZYNQ-7000开发之三】ZYNQ平台的HDMI驱动测试这篇文章，本次实验默认大家已经完成了【ZYNQ-7000开发之三】和【ZYNQ-7]]></description><category></category></item><item><title><![CDATA[【ZYNQ-7000开发之九】使用VDMA在PL和PS之间传输视频流数据]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50516181</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50516181</guid><author>RZJMPB</author><pubDate>Thu, 14 Jan 2016 11:35:02 +0800</pubDate><description><![CDATA[VDMA可以把AXI4-Stream 类型的视频流通过S2MM，写入到DDR3中，反之也可以通过MM2S读入到VDMA接口的外设中。通过内嵌FPGA逻辑分析仪进行观察数据。]]></description><category></category></item><item><title><![CDATA[ADI官方 HDMI IP core使用的问题]]></title><link>https://blog.youkuaiyun.com/RZJMPB/article/details/50487233</link><guid>https://blog.youkuaiyun.com/RZJMPB/article/details/50487233</guid><author>RZJMPB</author><pubDate>Sat, 09 Jan 2016 11:11:56 +0800</pubDate><description><![CDATA[在这个路径下：\hdl\library\axi_hdmi_tx 
打开axi_hdmi_tx_ip.tcl 
可以看到以下的代码：  adi_ip_create axi_hdmi_tx
  adi_ip_files axi_hdmi_tx [list \
  "$ad_hdl_dir/library/common/ad_axi_ip_constr.xdc" \
  "$ad_hdl_dir/libr]]></description><category></category></item></channel></rss>