Ubuntu16.04 + CAFFE-SSD + [cpu_only] + Kitti Training Summary

This training main reference: http://blog.csdn.net/jesse_mx/article/details/65634482

Thank jesse_mx, helped me a lot.

Pit One "OpenCV not installed successfully":

OpenCV not installed success will lead to a variety of inexplicable problems, such as: "No module named Cv2" problem.

However, the correct solution to the OPENCV installation problem is not a smooth way. Here are the OpenCV installation methods, which have been summarized by several failures individually:

Main references:

Install and test OpenCV http://blog.csdn.net/pengz0807/article/details/49915573 in Ubuntu

The OPENCV version used is 2.4.10

Where the CMake command should read:

cmake-d cmake_build_type=release-d cmake_install_prefix=/usr/local-d with_tbb=on-d install_python_examples=on-d Build_examples=on-d with_qt=on-d With_opengl=on-dwith_eigen=off.

　　　　

sudo SH ' echo "/usr/local/lib" >/etc/ld.so.conf.d/opencv.conf

To change to

sudo SH ' echo "/usr/local/lib" >/etc/ld.so.conf.d/opencv.conf

　　

Resolve the "No module named Cv2" issue:

sudo Install Opencv-python

PS: Install 2.x version of Do not need to download ippicv_linux_20151201.tgz this thing, 3.x version of the need

Pit two "compile Caffe":

Caffe compilation requires a lot of dependent libraries, so the installation process of a variety of wonderful problems is normal. Here is the Caffe installation:

Main reference URL: Reference site: http://blog.csdn.net/muzilinxi90/article/details/53673184

　　Note:git clone https://github.com/BVLC/caffe.git This is going to be slow .

which

 for  in $ (catdoesinstall done

To be replaced by:

 for  in $ (catdoessudoinstall done

　　It takes more than a few times to make sure that the complete installation is successful without any problems!

"Solution: Ubuntu no/usr/include/hdf5/serial/"

Step 1

In line 85th of the Makefile.config file, add/usr/include/hdf5/serial/to include_dirs, i.e. change the first line below to the second line of code.

Include_dirs: = $ (python_include)/usr/local/includeinclude_dirs:= $ (python_include)/usr/local/include/ usr/include/hdf5/serial/

Step 2

In line 173th of the makefile file, change Hdf5_hl and Hdf5 to Hdf5_serial_hl and hdf5_serial, i.e. change the first line of code below to the second line.

LIBRARIES += + glog gflags protobuf boost_system boost_filesystem m HDF5_SERIAL_HL hdf5_serial

　　

"FIX: No get_image_size tool, hint get_image_size not found"

　　 that's because you're using the original caffe and need to switch to the CAFFE-SSD branch :

Method One:

git clone https://github.com/weiliu89/caffe.gitcd caffe/git checkout SSD # Switch to branch SSD

Method Two:

Download GitHub's Zip package directly: https://GITHUB.COM/WEILIU89/CAFFE/TREE/SSD

Pit three "cpu_only":

Compiling caffe with Cpu_only mode will encounter more problems, so it is highly recommended to use GPU mode .

If a training terminal appears and an error is similar to the following:

Sgd_solver. CPP:106131000.001math_functions. cpp:+] Check failed:a <= B <0 vs-1.19209e-007>*** check failure stack t Race * * *.

Then you can find the corresponding CPP file and comment out the code similar to the following in the corresponding location.

// Check_le (A, b);

Then recompile Pycaffe and make all, and start training again.

"FIX: Training interrupted, insufficient memory"

Even if the value of batch_size is changed, the following error may occur on a machine with low configuration:

Check failed: *ptr host allocation of size184320000failedCheck failure Stack Trace: * * *    @ 0xb72472b5google::logmessage::fail () @0xb724957fGoogle::logmessage::sendtolog () @0xb7246deeGoogle::logmessage::flush () @0xb724a06dgoogle::logmessagefatal::~logmessagefatal () @0xb76190b9caffe::syncedmemory::mutable_cpu_data () @0xb73e8260Caffe::blob<>:: Mutable_cpu_data () @0xb7505c51Caffe::P oolinglayer<>:: Forward_cpu () @0xb7413e15Caffe::net<>:: Forwardfromto () @0xb7414062Caffe::net<>:: Forward () @0xb7637d51Caffe::solver<>:: Step () @0xb76388a9Caffe::solver<>:: Solve () @0x8053221train () @0x804eff8Main @0xb69e2637__libc_start_main @0X804F8CB(unknown)

　　　

At this point, you need to change the configuration .... There is no other way ...

Ubuntu16.04 + CAFFE-SSD + [cpu_only] + Kitti Training Summary