libcnn is a modular deep learning libraray, useful for robotics and computer vision. It is being developed by Robotics and Multi-perception Lab of City Univerisity of Hong Kong.
Here is a quick version to help you get started with the library.
The library has a light dependancy list:
- Eigen vesion 3, a C++ template library for linear algebra: matrices, vectors, numerical solvers, and related algorithms.
- boost version 1.49, peer-reviewed portable C++ source libraries.
The library was compiled on Ubuntu
mkdir build
cd build
cmake ..
make
sudo make install (optional)
We show an example of pixel-wise classification: scene-labelling.
The example used semantically-augmented make3d dataset. The following
assumes that you have already successfully done the above.
cd build
./train (training)
./validate (testing)
Here are several pixel-wise classification results of semantically-augmented make3d dataset
Please use github's issue tracker to report bugs