caffe-windows
caffe-windows copied to clipboard
happynear
Configure Caffe in one hour for Windows users.
Great work,thanks for sharing, I appreciate it! I am trying to apply caffe.binding project to encapsulate a DLL to extract face features. I am not very clear on how to...
Hi, happynear,when i call the function of TestFDDBPrecision in your MTCNN code with your caffe-windows. I have turned off the CUDNN, it also have the error.
Hello All, I am trying to get python layers working on the windows version of Caffe on visual studio. I am using a CPU only solution with everything out of...
风神，我在train.prototxt中定义了一个python的数据层，然后运行时出现这样的报错，在别人的linux运行是正常的，我看到问题里面也有好多人出现这样的问题。可是没人解决这个问题，这个问题究竟什么原因？怎么解决呢？
I followed the instructions. And compile both libcaffe and pycaffe thes two projects successful. And After I copy pycaffe/caffe into ../Lib/site-packages/ and import caffe in cmd. It says: from ._caffe...
https://github.com/happynear/caffe-windows/blob/2e9ade3075d86321342966ed0aa2961031a06daf/src/caffe/layers/inner_product_layer.cpp#L114-L117 @happynear 如题所示，貌似在处理weight归一化时只考虑了非transpose的情况(此时weight的维数为：NxK)，但是weight进行transpose的情况(此时weight的维数为：KxN)好像不能用相同的代码进行处理了。 weight归一化不应该是对每个(共N个，为输出类别数目)长度为K的weight向量进行归一化吗？？？
https://github.com/happynear/caffe-windows/blob/2e9ade3075d86321342966ed0aa2961031a06daf/src/caffe/layers/inner_product_layer.cpp#L141-L165 @happynear 你好，既然在Forward中weight进行了归一化：Y=WX/||W||，那么在Backward中weight的导数计算方式怎么跟没有进行归一化(Y=WX)时一样，都是bottom_data*top_diff (for transpose top_diff )或者top_diff *bottom_data(for no transpose top_diff)，这样计算是为了工程上处理方便，还是有别的特殊考虑？？
人脸验证年前试了一下，实话说，效果不是很理想，目前我也测试了很多公司的demo，貌似只有北京飞搜科技的demo最准，基本上能达到实用的地步，百度的也还行，学院派的那几个基本上都不行，具体名字就不说了。不知道影响准确度的原因到底是啥，数据还是算法，如何提高。我想做离线的人脸验证，从商务方面来说，用不了飞搜和百度的（原因比较多），只有自己做，本来想基于facenet做，后来测试发现效果也很不好，没思路了，郁闷ing！
Good Day, I have managed to build the caffe-windows lib under the release mode using VS2015 (with CPU only). I downloaded a VGG Deep Face Model (for caffe) via the...
