Caffea:Deep Learning Framework with Expression, Speed, and Modularity

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Caffe

本帖隐藏的内容

caffe-master.zip (8.37 MB)

2016-7-1 11:02:07 上传

Caffe is a deep learning framework made with expression, speed, and modularity in mind. It is developed by the Berkeley Vision and Learning Center (BVLC) and community contributors.

Check out the project site for all the details like

• DIY Deep Learning for Vision with Caffe
• Tutorial Documentation
• BVLC reference models and the community model zoo
• Installation instructions
  

and step-by-step examples.

Please join the caffe-users group or gitter chat to ask questions and talk about methods and models. Framework development discussions and thorough bug reports are collected on Issues.

Happy brewing!

License and Citation

Caffe is released under the BSD 2-Clause license. The BVLC reference models are released for unrestricted use.

Please cite Caffe in your publications if it helps your research:

@article{jia2014caffe,  Author = {Jia, Yangqing and Shelhamer, Evan and Donahue, Jeff and Karayev, Sergey and Long, Jonathan and Girshick, Ross and Guadarrama, Sergio and Darrell, Trevor},  Journal = {arXiv preprint arXiv:1408.5093},  Title = {Caffe: Convolutional Architecture for Fast Feature Embedding},  Year = {2014}}

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：modularity Expression Framework Learning earning expression framework developed reference learning

1. Caffe
   
2. 
   
3. Caffe is a library famous for its speed. The official project page is http://caffe.berkeleyvision.org/ and the GitHub page is https://github.com/BVLC/caffe. Similar to TensorFlow, Caffe has been developed mainly with C++, but it provides a Python and MATLAB API. In addition, what is unique to Caffe is that you don't need any programming experience, you just write the configuration or protocol files, that is .prototxt files, to perform experiments and research with deep learning. Here, we focus on the protocol-based approach.
   
4. 
   
5. Caffe is a very powerful library that enables quick model building, training, and testing; however, it's a bit difficult to install the library to get a lot of benefits from it. As you can see from the installation guide at http://caffe.berkeleyvision.org/installation.html, you need to install the following in advance:
   
6. 
   
7. CUDA
   
8. BLAS (ATLAS, MKL, or OpenBLAS)
   
9. OpenCV
   
10. Boost
    
11. Others: snappy, leveldb, gflags, glog, szip, lmdb, protobuf, and hdf5
    
12. Then, clone the repository from the GitHub page and create the Makefile.config file from Makefile.config.example. You may need Anaconda, a Python distribution, beforehand to run the make command. You can download this from https://www.continuum.io/downloads. After you run the make, make test, and make runtest commands (you may want to run the commands with a -jN option such as make -j4 or make -j8 to speed up the process) and pass the test, you'll see the power of Caffe. So, let's look at an example. Go to $CAFFE_ROOT, the path where you cloned the repository, and type the following commands:
    
13. 
    
14. $ ./data/mnist/get_mnist.sh
    
15. $ ./examples/mnist/train_lenet.sh
    
16. That's all you need to solve the standard MNIST classification problem with CNN. So, what happened here? When you have a look at train_lenet.sh, you will see the following:
    
17. 
    
18. #!/usr/bin/env sh
    
19. 
    
20. ./build/tools/caffe train --solver=examples/mnist/lenet_solver.prototxt
    
21. It simply runs the caffe command with the protocol file lenet_solver.prototxt. This file configures the hyper parameters of the model such as the learning rate and the momentum. The file also references the network configuration file, in this case, lenet_train_test.prototxt. You can define each layer with a JSON-like description:
    
22. 
    
23. layer {
    
24.   name: "conv1"
    
25.   type: "Convolution"
    
26.   bottom: "data"
    
27.   top: "conv1"
    
28.   param {
    
29.     lr_mult: 1
    
30.   }
    
31.   param {
    
32.     lr_mult: 2
    
33.   }
    
34.   convolution_param {
    
35.     num_output: 20
    
36.     kernel_size: 5
    
37.     stride: 1
    
38.     weight_filler {
    
39.       type: "xavier"
    
40.     }
    
41.     bias_filler {
    
42.       type: "constant"
    
43.     }
    
44.   }
    
45. }
    
46. So, basically, the protocol file is divided into two parts:
    
47. 
    
48. Net: This defines the detailed structure of the model and gives a description of each layer, hence whole neural networks
    
49. Solver: This defines the optimization settings such as the use of a CPU/GPU, the number of iterations, and the hyper parameters of the model such as the learning rate
    
50. Caffe can be a great tool when you need to apply deep learning to a large dataset with principal approaches.

复制代码

ahahahahhaha

ahahahahhaha

哈哈ｋｋｗｋｗｋｗｋ

感谢分享

感谢分享～