Theano+OpenCV+LIFT

2018-06-28

0.说明

LIFT: Learned Invariant Feature Points
作者开源代码：https://github.com/cvlab-epfl/LIFT

PS: 注意下载作者release的1.0版，修复了一些bug:.locks directory is now included in the repository by default。

1.安装最新OpenCV 3.x

略

2.安装Theano并配置

pip install theano==0.9.0

配置：(开源代码中三个主要部分需要GPU的支持)

sudo vim ~/.theanorc

[global]  
device = gpu0  
floatX = float32 

[nvcc]  
fastmath = True  

[cuda]    
root=/usr/local/cuda-8.0

3.安装其他依赖项

tar -zxf flufl.lock-2.4.1.tar.gz
sudo python setup.py install

2) 或

pip install flufl.lock==2.4.1

安装Lasagne:
下载安装包：https://github.com/Lasagne/Lasagne【(0.2.dev1)】

tar -zxf Lasagne.tar.gz
sudo python setup.py install

其他项：

pip install numpy parse h5py scipy

4.测试demo

# Build the shared library by
cd c-code/build
cmake -D CUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-8.0 ..
make

#To run the test program simply
./run.sh

5.其他问题

cmake报错：

Could NOT find CUDA: Found unsuitable version "9.1", but required is exact

version “8.0” (found /usr/local/cuda-9.1)

#需指定cuda版本
cmake -D CUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-8.0 ..