Skip to content

Latest commit

 

History

History
99 lines (67 loc) · 3.56 KB

README.md

File metadata and controls

99 lines (67 loc) · 3.56 KB
Raw

Jetson-TX2-Install-Caffe

Installing Caffe v1 official release from BVLC in embedded device Nvidia Jetson TX2 8GB with JetPack L4T 4.2 as updated as August 2019 rellease.

Official Caffe: https://caffe.berkeleyvision.org/

Getting Started:

For JetPack L4T v4.2 as of latest August 2019 JetPack release Ubuntu 18.04 comes with OpenCV v3.3 (No CUDA support), c++ (Ubuntu/Linaro 7.4.0-1ubuntu1~18.04.1) 7.4.0, Python 3.6 and CUDA 9.0 with 7.x cuDNN

Prerequisites

Dependencies for installation: sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev
libhdf5-dev libhdf5-serial-dev protobuf-compiler sudo apt-get install --no-install-recommends libboost-all-dev sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev sudo apt-get install libatlas-base-dev libopenblas-dev

Installing

In your terminal type git clone https://github.com/BVLC/caffe cd caffe cp Makefile.config Makefile.config.example (download Makefile.config and modify paths according to device and user name) make -j6 all make -j6 test make runtest mkdir -p ~/src #As of recommendation from jkjung in https://jkjung-avt.github.io/caffe-on-tx2/ I installed # updated package leveldb-0.20 ldconfig # distribute Caffe in the system cd ~/src wget https://pypi.python.org/packages/03/98/1521e7274cfbcc678e9640e242a62cbcd18743f9c5761179da165c940eac/leveldb-0.20.tar.gz tar xzvf leveldb-0.20.tar.gz cd leveldb-0.20 python3 setup.py build sudo python3 setup.py install pkgs=sed 's/[>=<].*$//' ~/caffe/python/requirements.txt for pkg in $pkgs; do sudo pip3 install $pkg; done cd ~/caffe #make caffev wrapper for Python3 --> pycaffe make pycaffe

to check that caffe is installed type .caffe/build/tools/caffe # Shows Caffe options to do train, test and time a actions

Running the tests

./caffe/build/tools/caffe time --gpu 0 --model caffe/examples/mnist/lenet.prototxt #Measure execution time of 1 iteration, you can also add number of iterations with "--iteration" ./caffe/build/tools/caffe train --solver --gpu 0 # Train your model ./caffe/build/tools/caffe train --gpu 0 --solver path_to_solver/solver.prototxt --snapshot path_to_snapshot/caffe_net_1_iter_1000.solverstate #Resume your training ./caffe/build/tools/caffe test —model path/to/deploy/deploy.prototxt —-weights path/to/trained/weights.caffemodel —-gpu 0 —iterations 100 # Benchmark of trained file for 100 iterations

To run automated tests of Caffe for classification go to https://caffe.berkeleyvision.org/gathered/examples/cpp_classification.html For more documentation visit: https://caffe.berkeleyvision.org

Deployment

To let Caffe use your Python3 scripts it is required to export PYTHONPATH and add the following line to your ".bashrc" export PYTHONPATH=/path/to/python3/script/:$PYTHONPATH Caffe package is imported in python scripts using: import caffe # Rememeber to install scikit-image v0.15 or higher and numpy v1.17 to avoid "no module named # skimage.io"

To use your C++ scripts with Caffe it is required a Cmake installation (2nd part of this guide coming soon) with CmakeLists.txt With this kind of installation Caffe package is imported in C++ scripts using: #include "caffe.hpp"

Built With

  • Caffe dependencies
  • scikit-image v0.15
  • leveldb-0.20

Contributing

https://jkjung-avt.github.io/caffe-on-tx2/ for installation of Caffe in Jetson TX2 with JetPack 3.3

Versioning

Version 1.0 Alpha

Authors

  • Wheslyx

License

Use this information at your discretion.

Acknowledgments

BVLC at http://caffe.berkeleyvision.org/ for developing this great framework "Caffe"