This caffe is adapted to meet the requirement of fine-grained aerial image recognition tasks (e.g. road and building segmentation and classification).
This repo is verified on bvlc/caffe docker image; just simply deploy this repo into /opt/caffe
of the docker container. The brief steps to compile this repo are
- In terminal, run
docker pull bvlc/caffe:cpu
- Start docker with a local folder mounted:
docker run -it --name caffe_test -v /path/to/local/folder:/dvol bvlc/caffe:cpu
- Copy the entire repo into
/path/to/local/folder
and then into/opt/caffe
overwritting folders. - Install libmatio:
sudo apt-get update && sudo apt-get install libmatio-dev
- In folder
/opt/caffe/build
, runcmake .. && make all -j4
and wait till it finishes - Test the installation using ...
Note: if you get error like "./dlcrfRF: error while loading shared libraries: libcaffe.so.1.0.0: cannot open shared object file: No such file or directory", try to set:
export LD_LIBRARY_PATH=/opt/caffe/build/lib:$LD_LIBRARY_PATH