Use OpenCV & OpenMP on OSX Yosemite 10.10.3 GNU GCC or LLVM/Clang ?

筆記文..

在遇到這問題之前，我已成功編譯且灌好歐噴屎Ｖ，直到昨天需要編譯一個使用了 OpenCV 和 OpenMP 的專案，問題就產生了！

我無法成功使用 gcc 在 osx 上編譯 OpenCV..所以之前是用系統預設的 clang，但 Yosemite 底下的 clang 似乎沒有支援 OpenMP 了，導致專案編譯時無法成功（即使灌了 OpenMP）

目前解法：

1. 重新安裝 gcc 和 OpenMP

brew uninstall gcc
brew install open-mpi

先移除原有的 gcc，brew install open-mpi 會安裝 gcc 和 openmp。

 
1. 重新編譯  OpenCV
修改 path to opencv-2.4.10/cmake/OpenCVCompilerOptions.cmake

set(OPENCV_EXTRA_FLAGS " -stdlib=libstdc++")
set(OPENCV_EXTRA_C_FLAGS " -stdlib=libstdc++")
set(OPENCV_EXTRA_CXX_FLAGS " -stdlib=libstdc++")
set(OPENCV_EXTRA_EXE_LINKER_FLAGS " -stdlib=libstdc++")

加上 -stdlib=libstdc++ 原因是現在 calng 預設使用 libc++，但 g++ 是 libstdc++，應該很清楚為何這麼做了！之後編譯專案時為了能用 OpemMP，選擇使用 g++ 那麼編譯 OpenCV 當然要選 g++ 用的 libstdc++ 囉！（改完後安裝方式應該都很清楚就不多說了，記得編譯使用 tbb 就不能用 OpenMP！我是選 tbb..）

 
1. 修改專案使用的預設編譯器（CMakeLists.txt）

set(CMAKE_C_COMPILER "/usr/local/Cellar/gcc/4.9.2_1/bin/gcc-4.9")
set(CMAKE_CXX_COMPILER "/usr/local/Cellar/gcc/4.9.2_1/bin/g++-4.9")

Install OpenNI2 on Ubuntu 14.04 with Xtion(PrimeSense)

First i've try OpenNI 2 SDK Binaries unfortunately it can't find my device so i install it like i did before.. 1. compile driver 2. install openni2

1. create OpenNI2 folder

mkdir OpenNI2 && cd OpenNI2

2. compile driver

git clone git@github.com:PrimeSense/Sensor.git
cd Sensor
cd Platform/Linux/CreateRedist
./RedistMaker
cd ../../..

cd Platform/Linux/Redist
cd Sensor-Bin-*
sudo ./install.sh

# back to OpenNI2 folder
cd ../../../../../

3. install openni2

# or you can get develop branch
# git clone -b develop git@github.com:occipital/OpenNI2.git
git clone  git@github.com:occipital/OpenNI2.git
cd OpenNI2 && make -j8 && cd ..

4. now plug in Xtion

cd OpenNI2/Bin/x64-Release && ./NiViewer
# will see the result like below press ESC to exit

Use minicom to connect Raspberry PI

1. Prepare
a. A USB to TTL Serial
b. Install minicom

sudo apt-get install minicom

2. Connect
a. connect TTL to GPIO like below



b. find the port on your PC

ls -l /dev/ttyUSB0

the output will like "crw-rw---- 1 root dialout...", means it's character device


sudo minicom -b 115200 -o -D /dev/ttyUSB0

Then you'll coneect to RPI !

Easily Stream your Camera on Ubuntu 14.04

1. Install dependent lib

sudo apt-get install imagemagick
sudo apt-get install libjpeg8-dev
sudo apt-get install libjpeg-dev
sudo apt-get install libv4l-dev

2. Download MJPG-Streamer and Install

svn checkout svn://svn.code.sf.net/p/mjpg-streamer/code/ mjpg-streamer-code
cd mjpg-streamer-code/mjpg-streamer
make
sudo make install

3. Start the Server!

mjpg_streamer -i "/usr/local/lib/input_uvc.so -f 15 -r 640x480" -o "/usr/local/lib/output_http.so -w /usr/local/www -p 8080"

Install caffe on ubuntu 14.04 with cuda-6.5.14

My device:

1. OS: ubuntu 14.04
2. linux kernel: 3.13.0-39
3. graphic card: nvidia gtx 660

(I can't promise that this way can do well on laptop which graphic card include nvidia Optimus mechanisum)

1. Install CUDA (include nvidia driver)
• Install build essential

sudo apt-get update && sudo apt-get install build-essential

• Download CUDA and Extract

wget http://developer.download.nvidia.com/compute/cuda/6_5/rel/installers/cuda_6.5.14_linux_64.run

chmod a+x cuda_6.5.14_linux_64.run

mkdir nvidia_installer

./cuda_6.5.14_linux_64.run -extract=`pwd`/nvidia_installer

• Press Ctrl-Alt-F1 and login
Stop the lightdm and remove current nvidia driver

sudo service lightdm stop
sudo apt-get remove --purge nvidia-*
# If it returns there are some process cause it can't be remove, i just kill them all and remove again.



2. Install

cd nvidia_installer
sudo ./NVIDIA-Linux-x86_64-340.29.run
sudo ./cuda-linux64-rel-6.5.14-18749181.run
sudo ./cuda-samples-linux-6.5.14-18745345.run
# restart lightdm and disable nouveaus
sudo service lightdm start

(I set every settings by default!)
3. Disable nouveau

gedit /etc/modprobe.d/blacklist-nouveau.conf

and add

blacklist nouveau
blacklist lbm-nouveau
options nouveau modeset=0
alias nouveau off
alias lbm-nouveau off

Reboot

update-initramfs -u
sudo reboot

4. Other necessary libs.
cudnn api download (Login and answer some questions to get the download permission.)

tar -xzvf cudnn-6.5-linux-R1.tgz
cd cudnn-6.5-linux-R1
sudo cp lib* /usr/local/cuda/lib64/
sudo cp cudnn.h /usr/local/cuda/include/

sudo cp lib* /usr/local/lib
sudo cp cudnn.h /usr/local/include
sudo ln -sf /usr/local/lib/libcudnn.so.6.5.18 /usr/local/lib/libcudnn.so.6.5

OpenBLAS

git clone https://github.com/xianyi/OpenBLAS
cd OpenBLAS
make
mkdir build
make install PREFIX=build

leveldb

git clone https://github.com/google/leveldb
cd leveldb
make
sudo cp --preserve=links libleveldb.* /usr/local/lib
sudo cp -r include/leveldb /usr/local/include/
sudo ldconfig

others

sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libboost-all-dev libhdf5-serial-dev libgflags-dev libgoogle-glog-dev liblmdb-dev protobuf-compiler

5. Install caffe and testing

git clone https://github.com/BVLC/caffe
cd caffe
cp Makefile.config.example Makefile.config
gedit Makefile.config

edit:
• USE_CUDNN := 1
• BLAS := open
• BLAS_INCLUDE := your-path-to-blas/OpenBLAS/build/include
• BLAS_LIB := your-path-to-blas/OpenBLAS/build/lib
install

#compile
make all
# test
make testexport LD_LIBRARY_PATH=~/Library/OpenBLAS/build/lib/
make runtest

Eclipse 更新 adt 22.6.2

不知怎麼，最近想更新 adt 22.6.2 ，卻一直出問題，最後是直接上官網下載 ADT-22.6.2.zip，但是更新完變成 sdk 不能更新了，最後是設定 Android SDK Manager 的 proxy，才解決，另外發現用宿舍網路還是有一部份被擋下來，所以改用吃到飽下載.....

Android SDK Manager => Settings

HTTP Proxy Server : www.google.com

HTTP Proxy Port : 80

Conversion to Dalvik format failed: Unable to execute dex: java.nio.BufferOverflowException. Check the Eclipse log for stack trace

如題，Import Android Project 之後 run，有時會出現這問題


[2013-11-10 17:54:15 - Dex Loader] Unable to execute dex: java.nio.BufferOverflowException. Check the Eclipse log for stack trace.
[2013-11-10 17:54:15 - testAndroid] Conversion to Dalvik format failed: Unable to execute dex: java.nio.BufferOverflowException. Check the Eclipse log for stack trace.


此時需要在專案上

點右鍵 → Build Path 將Android Dependecies Reamove 即可

Remove Android Dependecies