![](https://img-blog.csdnimg.cn/20201014180756927.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
Caffe
Hi_Panda_CRL
这个作者很懒,什么都没留下…
展开
-
caffe make runtest error(core dumped)Check failed: status == CUDNN_STATUS_SUCCESS (6 vs. 0)
参考了谷歌论坛=================================================================简单讲就是GPU的加速性能不够,CUDNN只支持CUDA Capability 3.0以上的GPU加速=================================================================转载 2015-06-04 15:02:22 · 4026 阅读 · 1 评论 -
Caffe专业术语
Blob:the blob is the standard array and unified memory interface for the framework. The details of blob describe how information is stored and communicated in and across layers and nets. A Blob is a原创 2016-04-14 08:38:55 · 605 阅读 · 0 评论 -
Caffe重新编译
从github获得的很多caffe源码都需要重新编译:流程:1、Makefile.config文件(cudnn、)2、修改Makefile文件(Open cv和MKL)修改mkl的路径 添加opencv包含库3、$ rm -rf ./build/* 清除build$ make all -j8$ make test -j8$ make runtest -j83原创 2016-05-20 18:03:22 · 4050 阅读 · 1 评论 -
Caffe --- blob code
两篇非常好的文章: http://blog.csdn.net/xizero00/article/details/50886829# http://www.cnblogs.com/yymn/articles/5341347.html原创 2016-07-23 13:02:21 · 888 阅读 · 0 评论 -
Blobs, Layers, and Nets: anatomy of a Caffe model
BlobsBlob作为Caffe的四大模块之一,负责完成CPU/GPU存储申请、同步和数据持久化映射。Caffe内部数据存储和通讯都是通过Blob来完成,Blob提供统一的存储操作接口,可用来保存训练数据、模型参数等。Blob 事实上是调用了SyncedMemory 类。SyncedMemory类封装了CPU/GPU内存申请、同步和释放等。所以SyncedMemory 完成了对内存的实际操作。原创 2016-07-21 15:39:26 · 899 阅读 · 0 评论 -
Caffe --- SyncedMemory
SyncedMemory类定义在syncedmem.hpp/cpp里, 主要负责caffe底层的内存管理.PS: Caffe的底层数据的切换(cpu模式和gpu模式),需要用到内存同步模块。其实个人觉得如果需要研究Blob,对于SyncedMemory的分析很重要内存分配与释放内存分配与释放由两个(不属于SyncedMemory类的)内联函数完成. 代码简单直观: 如果是CPU模式, 那么调用m原创 2016-07-25 09:59:49 · 677 阅读 · 0 评论 -
Caffe Solver
Solver scaffolds the optimization bookkeeping and creates the training network for learning and test network(s) for evaluation. iteratively optimizes by calling forward / backward and updating p原创 2016-07-21 15:46:39 · 492 阅读 · 0 评论 -
Caffe Log Visualization
1.Record your train/test log as a log fileTOOLS=./build/tools GLOG_logtostderr=0 GLOG_log_dir=deepid/deepid2/Log/ \ $TOOLS/caffe train \ --solver=deepid/deepid2/deepid_solver.prototxt 2.Parse原创 2016-09-28 14:28:41 · 953 阅读 · 0 评论 -
Caffe Layer Library
Convolution layer# convolutionlayer { name: "loss1/conv" type: "Convolution" bottom: "loss1/ave_pool" top: "loss1/conv" param { lr_mult: 1 decay_mult: 1 } param { lr_mult: 2原创 2017-03-14 15:22:33 · 4231 阅读 · 1 评论 -
deeplab script---python
import os,sys, subprocesssys.path.insert(0, os.getcwd()+'/python/my_script/')from tester import testerfrom trainer import trainerfrom crf_runner import crf_runner, grid_searchimport tools# MO原创 2016-06-01 07:53:21 · 1562 阅读 · 0 评论 -
deepLab
1.matio can't find HDF5 librarieschange file /densecrf/makefile as:g++ refine_pascal_v4/dense_inference.cpp util/Timer.h libDenseCRF.a$(CC) refine_pascal_v4/dense_inference.cpp -o prog_refine原创 2016-05-30 07:20:29 · 3271 阅读 · 0 评论 -
Caffe
1、下载:在官网(https://developer.nvidia.com/cudnn)下载,但是需要注册或者,在http://pan.baidu.com/s/1dDsD7XZ(密码:eqsz)下载,但是不保证是最新版2、安装:$ tar -xzvf cudnn-7.0-linux-x64-v3.0-rc.tgz$ cd cudnn-7.0-linux-x64-v3.0-r转载 2016-03-04 20:46:44 · 388 阅读 · 0 评论 -
Caffe Examples
1.ImageNet tutorialTrain and test "CaffeNet" on ImageNet data.原创 2016-03-14 08:34:34 · 2519 阅读 · 2 评论 -
LeNet: the MNIST Classification Model
name: "LeNet"layer { name: "mnist" type: "Data" top: "data" top: "label" include { phase: TRAIN } transform_param { scale: 0.00390625 } data_param { source: "examples/mn原创 2016-03-05 10:28:42 · 1207 阅读 · 0 评论 -
Train and Test LeNet on your own dataset
This post walk you through how to train LeNet (defined in examples/mnist/lenet.prototxt) using your own dataset( jpg images of the same size or preprocessed ) and then test using the trained LeNet.转载 2016-03-12 10:43:15 · 1128 阅读 · 0 评论 -
使用Python中draw_net.py画网络结构图
python python/draw_net.py /home/vr/Documents/Caffe/setup/caffe-master/examples/mnist/lenet.prototxt /home/vr/Documents/Caffe/setup/caffe-master/examples/mnist/lenet.png原创 2016-03-12 12:19:50 · 5041 阅读 · 0 评论 -
Caffe安装问题汇总
1.cannot open shared object file no such file or directory runtest error 127 caffeexport LD_LIBRARY_PATH=/usr/local/cuda-6.5/lib64:$LD_LIBRARY_PATHldconfig /usr/local/cuda-6.5/lib64/$ make clean$ m原创 2016-03-05 09:39:46 · 4163 阅读 · 0 评论 -
Caffe-Caffe Models
1.bvlc_reference_caffenet.caffemodel下载地址:http://dl.caffe.berkeleyvision.org/bvlc_reference_caffenet.caffemodel来源:caffe中ImageNet tutorial(Brew ImagesnNet)训练结果This model is the result of fol原创 2016-03-14 08:41:51 · 886 阅读 · 0 评论 -
bin to npy
import caffe % suppose caffe is already in the path of Pythonimport numpy as npimport sysblob = caffe.proto.caffe_pb2.BlobProto()data = open( 'data/ilsvrc12/imagenet_mean.binaryproto' , 'rb' ).re原创 2016-04-09 11:35:23 · 1493 阅读 · 0 评论 -
Caffe GNU error
*** Aborted at 1491880114 (unix time) try "date -d @1491880114" if you are using GNU date ***PC: @ 0x7fefd5f82cde (unknown)*** SIGSEGV (@0x0) received by PID 2769 (TID 0x7fefea08ca40) from PID 0;原创 2017-04-11 11:11:35 · 1232 阅读 · 2 评论