Нема описа

Alexander Freytag 061c83ef85 license issue пре 9 година
.gitignore 943926a51a first commit, supports loading of networks, image preparation, feat extraction for single image пре 10 година
README.md 061c83ef85 license issue пре 9 година
caffe_features_multiple_images.m ff55002afa support of filelists with relative filenames пре 9 година
caffe_features_single_image.m 76f292a096 safety check for normalization, no division by zero пре 9 година
caffe_load_network.m 9b6c843a8d adaptation towards recent caffe layout, made location globally known пре 9 година
caffe_prepare_image.m 943926a51a first commit, supports loading of networks, image preparation, feat extraction for single image пре 10 година
getFieldWithDefault.m c21c1976ce feature extraction supports bilinear pooling and logm transformation пре 10 година
initWorkspaceCaffeTools.m 4072de12c2 minor adaptation in init script пре 9 година

README.md

Summary

A small wrapper library to extract activations of deep neural networks using the popular Caffe toolbox.

Requirements / Dependencies

  • A compiled version of Caffe, tested with 2016-08-20 (hash 14643685a4 )
  • The source code was developed in Matlab 8.2

Installation

  • setup required paths and 3rd-party dependencies via initWorkspaceCaffeTools.m (if caffe is not found on your system, a possible code mirror is suggested)

Common errors

  • Do not forget to add caffe's dependencies (e.g., cuda, gflgs, protobuf, leveldb, lmdb, and mkl) to your LD_LIBRARY_PATH and PATH variable before starting Matlab

COPYRIGHT

This software is licensed under the non-commercial license Creative Commons Attribution 4.0 International License. For usage beyond the scope of this license, please contact Alex Freytag.