search

happynear / caffe-windows

Configure Caffe in one hour for Windows users.
Other
1.32k stars 650 forks source link

MTCNN #232

Open jianbaba opened 6 years ago

jianbaba commented 6 years ago

你好，请问您的mtcnn的（c++代码）vs文件夹下的前四个文件夹各实现什么功能？

happynear commented 6 years ago

应该挺清楚的吧，Mat是matlab warpper，Py是python wrapper，Test是测试，testPy是测试python wrapper的。

jianbaba commented 6 years ago

哦哦，懂了，您的det1-memory-stitch.prototxt + "det1.caffemodel"的作用是什么？还有里面的三个层的作用是？是您自己定义的层？我看原作者没有这个！ layer { name: "box“ type: "PredictBox" bottom: "Softmax" bottom: "conv4-2" top: "bb_map" top: "boudning_box" predict_box_param { positive_thresh: 0.6 stride: 2 receptive_field: 12 bbreg_exp: false } }

layer { name: "data" type: "MemoryData" top: "data" top: "label" memory_data_param{ batch_size: 1 height: 256 width: 256 channels: 3 } transform_param { mean_value: 127.5 scale: 0.0078125 } } layer { name: "stitch_data" type: "MemoryData" top: "stitch_data" top: "slabel" memory_data_param{ batch_size: 1 height: 123 width: 123 channels: 3 transpose: true } } @happynear

happynear commented 6 years ago

stitch的作用在这里写着： https://github.com/happynear/MTCNN_face_detection_alignment/tree/master/code/codes/vs

最好用我的caffe，我修改了很多地方。

jianbaba commented 6 years ago

@happynear https://github.com/happynear/caffe-windows/tree/ms 请问用您的caffe的意思是在这里下载并配置是么？能不能在linux下配置？要怎么做

happynear commented 6 years ago

linux的话还是有点困难的，主要是我把caffe封装成了一个动态库，而这个动态库在Linux下要自己写makefile来编译了。 https://github.com/happynear/caffe-windows/tree/ms/windows/caffe.binding

happynear commented 6 years ago

去翻翻别的实现吧，这个mtcnn有蛮多人复现和封装的。

jianbaba commented 6 years ago

@happynear 嗯嗯，好的谢谢您了，我就是要找一份c++实现，在linux下分装成静态库。因为我之前只用python调用caffe，c++基本小白，您写的c++代码对我也很有帮助，有助于我学习c++调用caffe，您还耐心帮我解答问题，真是太谢谢了