家家的专栏

NIPS2017的paper，但是文中介绍比较简单这里有详细解释：http://jalammar.github.io/illustrated-transformer/

cv::Mat gridimg(96,96,CV_8UC3,cv::Scalar(128,128,128));

Project-Out Cascaded Regression with an application to Face Alignment传统方法做的，效果还不错。

瞧

网络结构图：结果还不错。。。

结果图：

Generalized Video Deblurring for Dynamic ScenesKernel Fusion for Better Image DeblurringDeep Convolutional Neural Network for Image DeconvolutionLearning to Deblur (http://arxiv.

Devise: A deep visual-semantic embedding model (NIPS2013)Zero-Shot Learning by Convex Combination of Semantic Embeddings (http://arxiv.org/pdf/1312.5650.pdf)Mohammad Norouzi, Tomas Mikolov

1、object detection networks on convolutional feature maps2、Object detection via a multi-region & semantic segmentation-aware CNN model

code说明和下载地址：http://deeplearning.net/software/pylearn2/tutorial/index.html#tutorial

Deep Learning Tutorials:http://deeplearning.net/tutorial/

http://lsun.cs.princeton.edu/#organizers

cv::Mat gridimg(96,96,CV_8UC3,cv::Scalar(128,128,128));     for (int c=0; c    for (int h=0; h    for (int w=0; w              gridimg.at(h+hOffset, w+wOffset)[c] =                       m

下载地址：http://arxiv.org/pdf/1504.07889.pdf

VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION

好多年前学的东西，都忘记了，只能重新开始了。头文件：  #include 定义vector变量：  vector mean_cy;vector变量的赋值：  mean_cy.push_back(104);  mean_cy.push_back(117);  mean_cy.push_back(123);vector变量的使用：    mean_cy[0]

视频地址：http://www.ted.com/talks/fei_fei_li_how_we_re_teaching_computers_to_understand_pictures?language=enWhen a very young child looks at a picture, she can identify simple elements: "cat," "bo

框架：SegNet， a deep convolutional encoder-decoder architecture框架组成部分：A stack of encoders                       + a corresponding decoder stack                       + then feed into a softmax

数据库地址：http://www.cbsr.ia.ac.cn/english/CASIA-WebFace-Database.html

paper：        Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang. Learning a Deep Convolutional Network for Image Super-Resolution, in Proceedings of European Conference on Computer Vision (ECCV)

转自：http://www.cnblogs.com/huashiyiqike/p/3886670.html

示例：      printf("%d\n",mask[ph * pooled_width + pw]);

Paper： Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang, "Image Super-Resolution Using Deep Convolutional Networks", http://arxiv.org/abs/1501.0009 "For Creators", http://piapro.net/en_for_cr

blob。h函数修改：void save_to_file(const std::string& file_name);blob。cpp 的修改：template void Blob::save_to_file( const std::string& file_name){        FILE* fp = fopen(file_name.c_str(), "w

1. 大脑对视觉信息的处理是分层级的，低级脑区可能处理对边度，边缘什么的，高级脑区处理更抽象的比如人脸啊，房子啊，物体的运动之类的。信息被一层一层抽提出来往上传递进行处理。2. 大脑对视觉信息的处理也是并行的，不同的脑区提取出不同的信息干不同的活，有的负责处理这个物体是什么，有的负责处理这个物体是怎么动的。3. 脑区之间存在着广泛的联系，同时高级皮层对低级皮层也有很多的反馈投射。

VGG: http://www.robots.ox.ac.uk/~vgg/software/vgg_face/CMU-OpenFace: http://cmusatyalab.github.io/openface/

http://stackoverflow.com/questions/10526124/basic-build-issue-regarding-libs-pkg-config-and-opencv

http://serverfault.com/questions/201709/how-to-set-ld-library-path-in-ubuntu

文章主要贡献点：Flow-guided feature aggregation, an end-to-end framework for video object detection.Improve the per-frame features by aggregation of nearby features along the motion path, and thus impro

贡献点：           Construct CNNs which are capable of solving the optical flow estimation problem as a supervised learning task.            Train CNN end-to-end to learn predicting the optical flow f

贡献点:1. focus on the training data and show that the schedule of presenting data during training is very important. 2. develop a stacked architecture that includes warping of the second image with

1. Adversarial loss:    2. Cycle consistency loss  3. Full objective function:   Cycle GAN, 不仅让生成的图像和真实图像对抗学习，同时要求生成的图像能够decode出来原图像，这样不至于生成的图像太失真。文中有个例子，很好理解。就是不要要求中文翻译为英文，尽可能准确。还要求翻译...

转载自 楼燚(yì)航的blog，http://www.cnblogs.com/louyihang-loves-baiyan/https://github.com/YihangLou/fast-rcnn-train-another-dataset 这是我在github上修改的几个文件的链接，求星星啊，求星星啊（原谅我那么不要脸~~）在之前两篇文章中我介绍了怎么编译Fast RCNN，和

转自：http://blog.csdn.net/u011762313/article/details/47262549#配置pycaffe前言GPU版本正文安装依赖库一安装BLAS安装python安装matlab安装opencv安装依赖库二下载Caffe如果安装的是opencv30编译Caffe配置pycaffe配置matcaffe前言：按照Ca

Model:   Encoder-Recurrent-Decoder (ERD) model Task: for recognition and prediction of human body pose in videos and motion capture....         specially, motion capture generation, body pose labe

Deep hierarchies in the primate visual cortex: what can we learn for computer vision? TPAMI 2013 august.....

RGB-D图像其实是两幅图像：一个是普通的RGB三通道彩色图像，另一个是Depth图像。Depth图像类似于灰度图像，只是它的每个像素值是传感器距离物体的实际距离。通常RGB图像和Depth图像是配准的，因而像素点之间具有一对一的对应关系。利用Kinect可以采集RGB-D 图像。。。可以参考：Lai, K.; Bo, L.; Ren, X. & Fox, D. A large-sca

video address: http://techtalks.tv/talks/whats-wrong-with-deep-learning/61639/

Address: http://arxiv.org/list/cs.CV/recentBe updated everyday, check it everyday.