Object_Detector (Machine learning based object detector)

About

This is a machine learnng based object detector.
It uses dlib.
The project is built to demo face, hand, and logo detection.
This is a cocoapods project so open the project root in your terminal and run pod install. Once the command has finished open the Object Detector.xcworkspace file.
To run this project add the dependencies (see below) and compile dlib (see below).

Dependencies

DLIB can be downloaded from here. This was written with DLIB version 19.2.
This project requires that the iOS-Release version of dlib be installed into /usr/local/lib/DLIB/iOS. (see below for compile instructions)
This project also requires that libjpeg be installed. libjpeg can be downloaded via homebrew.
This project also requires that Open CV be installed. Open CV can be downloaded via homebrew.

DLIB Compile Instructions

This requires that cmake be installed. cmake can be downloaded via homebrew.
This requires that Xcode be downloaded.

To compile DLIB put the unzipped DLIB folder (dlib-xx.xx) into your desktop.

In the cmake file (CMakeLists.txt) in dlib-xx.xx/dlib comment out the line set (DLIB_USE_CUDA_STR "Disable this if you don't want to use NVIDIA CUDA" ) (cmake uses the # symbol to comment out lines).

NOTE: Cuda is the GPU library that dlib uses, this is currently not supported by iOS devices as it is NVIDIA. If you are using dlib on Mac OS and you have a NVIDIA GPU then you should leave this line uncommented.

Then run:

cd ~/Desktop/dlib-xx.xx/examples

mkdir build

cd build/

cmake -G Xcode ..

In dlib-xx.xx/examples/build there will now be a folder named dlib_build. In that folder is an Xcode prokect named dlib.xcodeproj. Open that project.

Select product->scheme->edit scheme and in the run section for each target set the build configuration to Release.

You will need to build the project for the architecure that you are targeting. In the build settings change the Base SDK for the desired target.

NOTE: If you build the project for Generic iOS device it will work on any iOS device but NOT the simulator and vice versa.

The products will appear in folders called e.g. Release or Release-iphoneos (see above for where to put the libraries you just built)

The contents of dlib-xx.xx/dlib needs to be copied to /usr/local/include/DLIB

Xcode Project Build Instructions

This is for setting up new projects to use DLIB

In the Build Settings for the project the following changes need to be made:

Under Build Options, set Enable Bitcode to NO
Under Linking add the following flags under Other Linker Flags:

/usr/lib/libpthread.dylib
/usr/lib/liblapack.dylib
/usr/local/lib/libjpeg.dylib
/usr/local/lib/libpng.dylib
-Wl,-search_paths_first
/usr/lib/libsqlite3.dylib
/usr/lib/libcblas.dylib
/usr/local/lib/DLIB/iOS/libdlib.a
-Wl,-headerpad_max_install_names

Under Search Paths, add the following paths to Header Search Paths:

/usr/local/include
/usr/local/include/DLIB/..
/usr/local/include/opencv

Also set User Header Maps to NO

Under Apple LLVM 7.0 - Code Generation, set Generate Debug Symbols to NO. Also set Optimization Level for Release to Fastest [-03]
Under ** Apple LLVM 7.0 - Custom Compiler Flags** add the following flags to Other C++ Flags for Debug:

-DDLIB_PNG_SUPPORT
-DDLIB_JPEG_SUPPORT
-DDLIB_USE_BLAS
-DDLIB_USE_LAPACK
-DENABLE_ASSERTS

and for Release:

-DDLIB_PNG_SUPPORT
-DDLIB_JPEG_SUPPORT
-DDLIB_USE_BLAS
-DDLIB_USE_LAPACK
-DNDEBUG

also add the following under Other Warning Flags:

-Wmost
-Wno-four-char-constants
-Wno-unknown-pragmas
$(inherited)

Under Apple LLVM X.X - Language - C++ set C++ Language Dialect to C++11[-std=c++11] and set C++ Standard Libary to Compiler Default.
Under Apple LLVM X.X - Preprocessing set Enable Foundation Assertions to YES. Also set Enable Strict Checking of objc_msgSend Calls to NO Also add the following flag under Preprocessor Macros: 'CMAKE_INTDIR="$(CONFIGURATION)$(EFFECTIVE_PLATFORM_NAME)"'
Ensure that all code files in the project are either C++ (.cpp) (.h) (.hpp) or Objctive-C++ (.mm) (.h) or C (.c) (.h)
Set the product scheme under build configuration to Release (see above)
Finnally add a copy of libdlib.a to the project (drag and drop). And build!

NOTE: With Xcode 8 there have been issues with library not found for -dlib build errors. To fix this: in the Library Search Paths remove any reference to the project directory and instead add a reference to /usr/local/lib where you should have a copy of the dlib library you built earlier.

Other Notes

This Folder also includes binaries, imgtrain and imglab. imgtrain is a command line app for creating new object detectors. imglab is a command line app for preparing images for imgtrain. See the README for each binary.