GSOC17 - Facemark API (#1257)
* Initial commit of facemark API
Initial structure of the facemark API and AAM header
* make training function as virtual
* Add: dataset parser
* Bug fix: clear the container before add points
* Add: AAM training - procrustes analysis
* Add AAM model
* Added training function for AAM
* Building bot fixes: remove training overload, explicit cast to float for atof
* + add dependency: imgcodecs
* Build bot fixes: add imgproc.hpp and type casting
* Building bot fix: type casting
* fixing the AAM training to match with Matlab version
fewer model parameters, change the image warp method, change the feature extraction method
* add: AAM fitting
added several functionalities for fitting
* fix warings
* Add: transformation for the initial fitting
* add sample file for aam implementation
* fix warning
* Add LFB Header
* loadTrainingData: Throw an error message if file not exist
* add: LBF prepare training data
* add: data augmentation
* change to double
* add: getMeanShape
* shuffling the dataset and parameters initialization
* add: initial structure of LBF class
* add: getDeltaShapes
Difference between the current shape and the desired shape
* add: random forest training
* generate lbf features
* global regression
* save training data
* fix the parameter initialization
* set the default parameters
* add: initial version of lbf sample
* update the current shape
* compute error
* add: prediction function
* fix some warnings
* fitting function
the result is mis-aligned, shuould be double checked
* add: fitting in the demo
* add dependencies
* Add: tutorial
* add: load model
* fixing training
* use user defined face detector
* Documents, tests, and samples
* Allow custom parameters
* Cleaning up
* Custom parameters for default detector, training, and get custom data
* AAM scales
* minor fixes , update the opencv_extra files
* change path to lbp cascade
* face: avoid memory leaks
* utilize the filestorage for the model, fixing some minor issues
* remove the liblinear dependency
* fix the aam test, avoiding to write any files
* use RNG and changes the test files
- broken bindings generator
- broken conversion functions (uses non-supported cv::Mat types)
- other problems like missing convertion functions for Moments and other structs
* kcf use float data type rather than double.
In our practice, float is good enough and could get better performance.
With this patch, one of my benchmark could get about 20% performance gain.
Signed-off-by: Zhigang Gong <zhigang.gong@intel.com>
* Offload transpose matrix multiplication to ocl.
The matrix multiplication in updateProjectMatrix is one of the
hotspot. And because of the matrix shape is special, say the
m is very short but the n is very large. The GEMM implementation
in neither the clBLAS nor the in trunk implementation are very
inefficient, I implement an standalone transpose matrix mulplication
kernel here. It can get about 10% performance gain on Intel
desktop platform or 20% performance gain on a braswell platform.
And in the mean time, the CPU utilization will be lower.
Signed-off-by: Zhigang Gong <zhigang.gong@intel.com>
* Add verification code for kcf ocl transpose mm kernel.
Signed-off-by: Zhigang Gong <zhigang.gong@linux.intel.com>
* tracking: show FPS in traker sample
* tracking: fix MSVC warnings in KCF
* tracking: move OCL kernel initialization to constructor in KCF
noise control implementation ended
noise control errors fixed (part I)
noise control errors fixed (ended)
Perf Test for staticSaliencySpectralResidual and objectnessBING added
improvements and bugs fixed
saliency perf test added
matrices optimization
Where possible, CV_8U matrices have been used in place of CV_32F
new sorting algorithm implemented (first part)
new sorting algorithm implemented (second part)
new sorting algorithm implemented (finished)
Alexander Alekhin
kurnianggoro
berak
LaurentBerger
cDc
Vladislav Sovrasov
Vitaly Tuzov
Maksim Shabunin
Pavel Rojtberg
Hamdi Sahloul
Alexander Alekhin
Gregor Kovalčík
atinfinity
Andrew Hundt
jaco