Caffe to generate your own files Lmdb

Reference website:

Http://www.cnblogs.com/darkknightzh/p/5909121.html (under Linux)

Http://www.mamicode.com/info-detail-1338521.html (under Windows)

Http://www.cnblogs.com/denny402/p/5082341.html (the simplest and most straightforward example)

http://www.2cto.com/kf/201607/527860.html (mnist data format and Lmdb conversion Code analysis)

HTTP://NBVIEWER.JUPYTER.ORG/GITHUB/BVLC/CAFFE/BLOB/MASTER/EXAMPLES/00-CLASSIFICATION.IPYNB (official website tutorial python do classification)

Caffe to generate your own files Lmdb

1. Build txt List of picture file lists

http://blog.csdn.net/sinat_30071459/article/details/51613304 (label making)

1. called convert_imageset.sh to convert

To create an executable file Create_imagenet_train.bat

2. .. \.. \build\x64\release\convert_imageset.exe
4. --resize_height=256
6. --resize_width=256 # Zoom
8. --shuffle # Scramble Order
10. F:\ImageDatabase\GoogleImage\GoogleImage-small\ # Image Path
12. D:\Dev\caffe-master\examples\googleimages\train.txt # Label Path
14. F:\ImageDatabase\GoogleImage\lmdb\googleimagessmall_train_lmdb # destination path, not present
18. Pause # Pause

similarly Create_imagenet_val.bat

1. Calculate mean File

Create a batch file Compute_mean.bat

1. .. \.. \build\x64\release\compute_image_mean.exe
2. F:\ImageDatabase\GoogleImage\lmdb\googleimagessmall_train_lmdb #train Lmdb Path
3. D:\Dev\caffe-master\examples\googleimages\mean.binaryproto # Output Path
5. Pause

Iv. List of documents

will be Lmdb documents, mean documents, label put the files in the right place .

Configure the path in the network configuration file to train

Caffe to generate your own files Lmdb