ZF/OpenFace
1
0
Fork 0
mirror of https://gitcode.com/gh_mirrors/ope/OpenFace.git synced 2026-08-01 00:17:45 +00:00
Code Issues Packages Projects Releases Wiki Activity

Full face model.

Browse Source
This commit is contained in:
Tadas Baltrusaitis
2016-06-03 16:53:27 -04:00
parent b75f12f078
commit 65e91da3a6
3 changed files with 78 additions and 41 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
exe/FeatureExtraction/FeatureExtraction.cpp
View File

@@ -153,7 +153,7 @@ void create_directory(string output_path)
}
void get_output_feature_params(vector<string> &output_similarity_aligned, vector<string> &output_hog_aligned_files, double &similarity_scale,
int &similarity_size, bool &grayscale, bool &rigid, bool& verbose,
int &similarity_size, bool &grayscale, bool &rigid, bool& verbose, bool& dynamic,
bool &output_2D_landmarks, bool &output_3D_landmarks, bool &output_model_params, bool &output_pose, bool &output_AUs, bool &output_gaze,
vector<string> &arguments);
@@ -235,7 +235,7 @@ void outputAllFeatures(std::ofstream* output_file, bool output_2D_landmarks, boo
cv::Point3f gazeDirection0, cv::Point3f gazeDirection1, const cv::Vec6d& pose_estimate, double fx, double fy, double cx, double cy,
const FaceAnalysis::FaceAnalyser& face_analyser);
void post_process_output_file(FaceAnalysis::FaceAnalyser& face_analyser, string output_file);
void post_process_output_file(FaceAnalysis::FaceAnalyser& face_analyser, string output_file, bool dynamic);
int main (int argc, char **argv)
@@ -306,6 +306,7 @@ int main (int argc, char **argv)
bool grayscale = false;
bool video_output = false;
bool rigid = false;
bool dynamic = true; // Indicates if a dynamic AU model should be used (dynamic is useful if the video is long enough to include neutral expressions)
int num_hog_rows;
int num_hog_cols;
@@ -318,7 +319,7 @@ int main (int argc, char **argv)
bool output_AUs = true;
bool output_gaze = true;
get_output_feature_params(output_similarity_align, output_hog_align_files, sim_scale, sim_size, grayscale, rigid, verbose,
get_output_feature_params(output_similarity_align, output_hog_align_files, sim_scale, sim_size, grayscale, rigid, verbose, dynamic,
output_2D_landmarks, output_3D_landmarks, output_model_params, output_pose, output_AUs, output_gaze, arguments);
// Used for image masking
@@ -359,13 +360,24 @@ int main (int argc, char **argv)
int curr_img = -1;
string au_loc;
if(boost::filesystem::exists(path("AU_predictors/AU_all_best.txt")))
string au_loc_local;
if (dynamic)
{
au_loc = "AU_predictors/AU_all_best.txt";
au_loc_local = "AU_predictors/AU_all_best.txt";
}
else
{
path loc = path(arguments[0]).parent_path() / "AU_predictors/AU_all_best.txt";
au_loc_local = "AU_predictors/AU_all_static.txt";
}
if(boost::filesystem::exists(path(au_loc_local)))
{
au_loc = au_loc_local;
}
else
{
path loc = path(arguments[0]).parent_path() / au_loc_local;
if(exists(loc))
{
@@ -684,16 +696,10 @@ int main (int argc, char **argv)
output_file.close();
if(output_files.size() > 0)
if(output_files.size() > 0 && output_AUs)
{
// If the video is long enough post-process it for AUs
if (output_AUs && frame_count > 1000)
{
cout << "Postprocessing the Action Unit predictions" << endl;
post_process_output_file(face_analyser, output_files[f_n]);
}
cout << "Postprocessing the Action Unit predictions" << endl;
post_process_output_file(face_analyser, output_files[f_n], dynamic);
}
// Reset the models for the next video
face_analyser.Reset();
@@ -718,7 +724,7 @@ int main (int argc, char **argv)
}
// Allows for post processing of the AU signal
void post_process_output_file(FaceAnalysis::FaceAnalyser& face_analyser, string output_file)
void post_process_output_file(FaceAnalysis::FaceAnalyser& face_analyser, string output_file, bool dynamic)
{
vector<double> certainties;
@@ -728,8 +734,8 @@ void post_process_output_file(FaceAnalysis::FaceAnalyser& face_analyser, string
vector<std::pair<std::string, vector<double>>> predictions_class;
// Construct the new values to overwrite the output file with
face_analyser.ExtractAllPredictionsOfflineReg(predictions_reg, certainties, successes, timestamps);
face_analyser.ExtractAllPredictionsOfflineClass(predictions_class, certainties, successes, timestamps);
face_analyser.ExtractAllPredictionsOfflineReg(predictions_reg, certainties, successes, timestamps, dynamic);
face_analyser.ExtractAllPredictionsOfflineClass(predictions_class, certainties, successes, timestamps, dynamic);
int num_class = predictions_class.size();
int num_reg = predictions_reg.size();
@@ -1024,7 +1030,7 @@ void outputAllFeatures(std::ofstream* output_file, bool output_2D_landmarks, boo
void get_output_feature_params(vector<string> &output_similarity_aligned, vector<string> &output_hog_aligned_files, double &similarity_scale,
int &similarity_size, bool &grayscale, bool &rigid, bool& verbose,
int &similarity_size, bool &grayscale, bool &rigid, bool& verbose, bool& dynamic,
bool &output_2D_landmarks, bool &output_3D_landmarks, bool &output_model_params, bool &output_pose, bool &output_AUs, bool &output_gaze,
vector<string> &arguments)
{
@@ -1041,6 +1047,9 @@ void get_output_feature_params(vector<string> &output_similarity_aligned, vector
string input_root = "";
string output_root = "";
// By default the model is dynamic
dynamic = true;
// First check if there is a root argument (so that videos and outputs could be defined more easilly)
for (size_t i = 0; i < arguments.size(); ++i)
{
@@ -1088,6 +1097,10 @@ void get_output_feature_params(vector<string> &output_similarity_aligned, vector
{
rigid = true;
}
else if (arguments[i].compare("-au_static") == 0)
{
dynamic = false;
}
else if (arguments[i].compare("-g") == 0)
{
grayscale = true;