编程语言,算法相关技术专家
Recurrent Models of Visual Attention Google DeepMind 模拟人类看东西的方式,我们并非将目光放在整张图像上,尽管有时候会从总体上对目标进行把握,但是也是将目光按照某种次序(例如,从上倒下,从左到右等等)在图像上进行扫描,然后从一个区域转移到另一个区域。
解决ssh的"Write failed: Broken pipe"问题 问题场景 服务器环境:阿里云 Linux CentOS 主机 客户端:Mac OSX Terminal 问题现象 用 ssh 命令连接服务器之后,如果一段时间不操作,再次进入 Terminal 时会有一段时间没有响应,然后就出现错误提示: Write failed: Broken pipe 只能重新用 ssh 命令进行连接。
Localizing by Describing: Attribute-Guided Attention Localization for Fine-Grained Recognition Baidu Research 本文主要是将part描述利用起来,协助进行part定位,针对每一个定位好的part,再进行每一个part对应属性的识别。
转载自:维基百科 蒙特卡洛方法 https://zh.wikipedia.org/wiki/%E8%92%99%E5%9C%B0%E5%8D%A1%E7%BE%85%E6%96%B9%E6%B3%95 蒙特卡洛方法[编辑] 维基百科,自由的百科全书 蒙特卡洛方法(英语:Monte Carlo method),也称统计模拟方法,是二十世纪四十年代中期由于科学技术的发展和电子计算机的发明,而被提出的一种以概率统计理论为指导的一类非常重要的数值计算方法。
Fully Convolutional Attention Localization Networks: Efficient Attention Localization for Fine-Grained Recognition 细粒度的识别(Fine-grained recognition)的挑战性主要来自于 类内差异(inter-class differences)在细粒度类别中通常是局部的,细微的;类间差异(intra-class differences)由于姿态的变换而导致很大。
Wide Residual Networks
浅谈 Hellinger Distance 2016.05.24 最近在看 Hellinger Distance(海林格距离), 平时看多了欧式距离,马氏距离等等,貌似介绍这个的材料不是很多,例如:维基百科上的一些,和 下列这个链接上的pdf材料:http://www.tcs.tifr.res.in/~prahladh/teaching/2011-12/comm/lectures/l12.pdf 大家凑合着看看吧,我这里简要的介绍下,有理解不对的地方,烦劳指点下。
MatchNet: Unifying Feature and Metric Learning for Patch-Based Matching CVPR 2015 本来都写到一半了,突然笔记本死机了,泪崩!好吧,重新写!本文提出了一种联合的学习patch表示的一个深度网络 和 鲁棒的特征比较的网络结构。
Compute Mean Value of Train and Test Dataset of Caltech-256 dataset in matlab code clc;imPath = '/home/wangxiao/Downloads/Link to caltech_256_...
SoftmaxLayer and SoftmaxwithLossLayer 代码解读 Wang Xiao 先来看看 SoftmaxWithLoss 在prototext文件中的定义: 1 layer { 2 name: "loss" 3 type: "SoftmaxW...
1. caffe matlab 接口提供了提取feature的脚本,但是由于中间要对这些图像进行RGB ---> BGR 的变换,卧槽,灰度图没有三通道啊?怎么破?从上午就在纠结怎么会跑着跑着程序就报错了,尼玛,坑啊。
Tutorial: Triplet Loss Layer Design for CNN Xiao Wang 2016.05.02 Triplet Loss Layer could be a trick for further improving the accuracy of CNN.
import random from urllib import urlopen import sys WORD_URL = "http://learncodethehardway.org/words.
1 # this will not be printed in python ! 2 print "I could have code like this." # and the comment after is ignored 3 4 # You can also us...
Active Object Localization with Deep Reinforcement Learning ICCV 2015 最近Deep Reinforcement Learning算是火了一把,在Google Deep Mind的主页上,更是许多关于此的pape...
如何给图像添加矩形框?以及添加想要输入的文字 ? 案例程序,如下所示: clc; close all; clear all;image = imread('/home/wangxiao/Pictures/image/gamersky_022origin_043_20147261368D.
Decoupled Deep Neural Network for Semi-supervised Semantic Segmentation xx
Feature visualization from ipython notebook Wang Xiao 1. install anaconda2 from: https://www.
1. 首先是提取 训练日志文件; 2. 然后是matlab代码: clear all; close all; clc; log_file = '/home/wangxiao/Downloads/43_attribute_baseline.
How to change the AlexNet into FCNs ? FCNs is a network that only contain convolution layers and no fc layer at all.
Instance-aware Semantic Segmentation via Multi-task Network Cascades Jifeng Dai Kaiming He Jian Sun 本文的出发点是做Instance-aware Semantic Segmentati...
Pedestrian Detection aided by Deep Learning Semantic Tasks CVPR 2015 本文考虑将语义任务(即:行人属性和场景属性)和行人检测相结合,以语义信息协助进行行人检测。
。。。
I0415 15:03:37.603461 27311 solver.cpp:42] Solver scaffolding done.I0415 15:03:37.603549 27311 solver.
Caffe + CUDA8.0 + CuDNNv5.1 + OpenCV3.1 + Ubuntu14.04 配置参考文献 ---- Wang Xiao AnHui University CVPR Group 2017-05-27 Warning: Please make ...
1. when compile /home/wangxiao/NVIDIA-CUDA-7.5 SAMPLES, it warning: gcc version larger than 4.9 not supported, so:old verson of gcc and g++ are needed: sudo apt-get install gcc-4.
NVCC src/caffe/solvers/adam_solver.cuIn file included from /usr/local/cuda/include/cuda_runtime.h:76:0, from :0:/usr/local/cuda/include/host_config.
如何给caffe添加新的layer ? 初学caffe难免会遇到这个问题,网上搜来一段看似经典的话, 但是问题来了,貌似新版的caffe并没有上面提到的vision_layer:
I0327 20:24:22.966171 20521 net.cpp:849] Copying source layer drop7I0327 20:24:22.966179 20521 net.cpp:849] Copying source layer fc8_I0327 20:24:22.
test code for PETA datasets .... 1 #ifdef WITH_PYTHON_LAYER 2 #include "boost/python.hpp" 3 namespace bp = boost::python; 4 #endif 5 ...
%% Rank the entropy_loss % for iiii = 1:size(Group_age, 1) % count_1 = 0 ;% tmp = Group_age(iiii, 1);% for jjj = 1:size(Group_age, 1)...
function script_faster_rcnn_demo() close all; clc; clear mex; clear is_valid_handle; % to clear init_key run(fullfile(filepart...
Check failed: FLAGS_weights.size() > 0 (0 vs. 0) Need model weights to score. 出现这个错误,但是我记得昨天还好好的,网上搜了也没有答案,后来仔细检查才发现,原来存放 .caffemodel 的文件名字 中间有空格!!! 把文件夹路径上的名字去掉,果断就可以了。
// write the predicted score into txt files ofstream file("/home/wangxiao/Downloads/caffe-master/wangxiao/bvlc_alexnet/predict_score.
1. Active Query Driven by Uncertainty and Diversity for Incremental Multi-Label Learning The key task in active learning is to design a selection ...
1 clc; clear all; close all; 2 image_path = '/media/wangxiao/Elements/image_segment_backup/'; 3 savePath = '/media/wangxiao/Seagate/wangxiao/...
wangxiao@wangxiao-GTX980:~/Downloads/caffe-master$ make -j8find: `wangxiao/bvlc_alexnet/spl': No such file or directoryfind: `caffemodel': No such file or directoryfind: `wangxiao/bvlc_alexnet/0.
I0221 21:47:41.826748 6797 solver.cpp:259] Train net output #0: loss = 0.00413362 (* 1 = 0.00413362 loss)I0221 21:47:41.
1 load fisheriris 2 data = [meas(:,1), meas(:,2)]; 3 groups = ismember(species,'setosa'); 4 [train, test] = crossvalind('holdOut',groups); 5 cp ...
I0219 14:48:40.965386 31108 net.cpp:76] Memory required for data: 0I0219 14:48:40.965517 31108 layer_factory.
when I compile caffe file :.build_debug/lib/libcaffe.so: undefined reference to `cv::imread(cv::String const&, int)'.
1 #include // for snprintf 2 #include 3 #include 4 5 #include "boost/algorithm/string.
------------------------------------------------------------------------------------------------------------------------------------------------------...
name: "VGG_ILSVRC_19_layer"layer { name: "data" type: "ImageData" top: "data" top: "label" include { phase: TRAIN } image_data_param { batch_size: 12 source: ".
最近跑深度学习,提出的feature是4096维的,放到我们的程序里,跑得很慢,很慢。。。。 于是,一怒之下,就给他降维处理了,但是matlab 自带的什么pca( ), princomp( )函数,搞不清楚怎么用的,表示不大明白,下了一个软件包: 名字:Matlab Toolbox for Dimensionality Reduction 链接:http://lvdmaaten.
利用caffe生成 lmdb 格式的文件,并对网络进行FineTuning 数据的组织格式为: 首先,所需要的脚本指令路径为: /home/wangxiao/Downloads/caffe-master/examples/imagenet/ 其中,生成lmdb的文件为: create_imagenet.sh 接下来的主要任务就是修改自己的data的存放路径了。
1. 判断是否存在指定的video_name, 若不存在,则在给定save_path下,新建一个video_name文件夹:1 sec_path = [save_path, video_name, '/']; 2 if ~exist(sec_path,'file') 3 mkdir([save_path, video_name]); 4 end2. 将gpuArray转换为matlab可以保存的格式,如:uint8 等。
DECOLOR: Moving Object Detection by Detecting Contiguous Outliers in the Low-Rank Representation Xiaowei Zhou et al.
Inferring Analogous Attributes CVPR 2014 Chao-Yeh Chen and Kristen Grauman Abstract: The appearance of an attribute can vary considerably from class to class (e.
OpenCV CommandLineParser 的用法 去百度了一下,关键字:OpenCV CommandLineParser 发现,最多的讲解是:opencv源码解析之(5):CommandLineParser类的简单理解 链接:http://www.