diff --git a/lib/local/LandmarkDetector/include/LandmarkDetectionValidator.h b/lib/local/LandmarkDetector/include/LandmarkDetectionValidator.h
index 89a9ef2f..b64af002 100644
--- a/lib/local/LandmarkDetector/include/LandmarkDetectionValidator.h
+++ b/lib/local/LandmarkDetector/include/LandmarkDetectionValidator.h
@@ -74,9 +74,9 @@ namespace LandmarkDetector
 {
 //===========================================================================
 //
-// Checking if landmark detection was successful using an SVR regressor
+// Checking if landmark detection was successful using a CNN
 // Using multiple validators trained add different views
-// The regressor outputs -1 for ideal alignment and 1 for worst alignment
+// The regressor outputs 1 for ideal alignment and 0 for worst alignment
 //===========================================================================
 class DetectionValidator
 {
diff --git a/lib/local/LandmarkDetector/include/LandmarkDetectorParameters.h b/lib/local/LandmarkDetector/include/LandmarkDetectorParameters.h
index 58c84d87..7fa17a5f 100644
--- a/lib/local/LandmarkDetector/include/LandmarkDetectorParameters.h
+++ b/lib/local/LandmarkDetector/include/LandmarkDetectorParameters.h
@@ -55,7 +55,7 @@ struct FaceModelParameters
 	// Should face validation be done
 	bool validate_detections;
 
-	// Landmark detection validator boundary for correct detection, the regressor output -1 (perfect alignment) 1 (bad alignment), 
+	// Landmark detection validator boundary for correct detection, the regressor output 1 (perfect alignment) 0 (bad alignment), 
 	float validation_boundary;
 
 	// Used when tracking is going well
@@ -104,9 +104,6 @@ struct FaceModelParameters
 	// Should the parameters be refined for different scales
 	bool refine_parameters;
 
-	// Using the brand new and experimental gaze tracker
-	bool track_gaze;
-
 	FaceModelParameters();
 
 	FaceModelParameters(vector<string> &arguments);
diff --git a/lib/local/LandmarkDetector/src/LandmarkDetectionValidator.cpp b/lib/local/LandmarkDetector/src/LandmarkDetectionValidator.cpp
index 479733ad..3cc400f4 100644
--- a/lib/local/LandmarkDetector/src/LandmarkDetectionValidator.cpp
+++ b/lib/local/LandmarkDetector/src/LandmarkDetectionValidator.cpp
@@ -377,6 +377,9 @@ float DetectionValidator::Check(const cv::Vec3d& orientation, const cv::Mat_<uch
 	// The actual validation step
 	double dec = CheckCNN(warped, id);
 
+	// Convert it to a more interpretable signal (0 low confidence, 1 high confidence)
+	dec = 0.5 * (1.0 - dec);
+
 	return (float)dec;
 }
 
diff --git a/lib/local/LandmarkDetector/src/LandmarkDetectorModel.cpp b/lib/local/LandmarkDetector/src/LandmarkDetectorModel.cpp
index e5deda17..3f5d7de5 100644
--- a/lib/local/LandmarkDetector/src/LandmarkDetectorModel.cpp
+++ b/lib/local/LandmarkDetector/src/LandmarkDetectorModel.cpp
@@ -590,8 +590,7 @@ bool CLNF::DetectLandmarks(const cv::Mat_<uchar> &image, FaceModelParameters& pa
 		{
 			// Only do the synthetic eye models if we're doing gaze
 			if (!((hierarchical_model_names[part_model].compare("right_eye_28") == 0 ||
-			hierarchical_model_names[part_model].compare("left_eye_28") == 0)
-			&& !params.track_gaze))
+			hierarchical_model_names[part_model].compare("left_eye_28") == 0)))
 			{
 
 				int n_part_points = hierarchical_models[part_model].pdm.NumberOfPoints();
@@ -638,8 +637,7 @@ bool CLNF::DetectLandmarks(const cv::Mat_<uchar> &image, FaceModelParameters& pa
 				vector<pair<int, int>> mappings = this->hierarchical_mapping[part_model];
 
 				if (!((hierarchical_model_names[part_model].compare("right_eye_28") == 0 ||
-					hierarchical_model_names[part_model].compare("left_eye_28") == 0)
-					&& !params.track_gaze))
+					hierarchical_model_names[part_model].compare("left_eye_28") == 0)))
 				{
 					// Reincorporate the models into main tracker
 					for (size_t mapping_ind = 0; mapping_ind < mappings.size(); ++mapping_ind)
@@ -663,7 +661,7 @@ bool CLNF::DetectLandmarks(const cv::Mat_<uchar> &image, FaceModelParameters& pa
 
 		detection_certainty = landmark_validator.Check(orientation, image, detected_landmarks);
 
-		detection_success = detection_certainty < params.validation_boundary;
+		detection_success = detection_certainty > params.validation_boundary;
 	}
 	else
 	{
diff --git a/lib/local/LandmarkDetector/src/LandmarkDetectorParameters.cpp b/lib/local/LandmarkDetector/src/LandmarkDetectorParameters.cpp
index cb263598..42f02f24 100644
--- a/lib/local/LandmarkDetector/src/LandmarkDetectorParameters.cpp
+++ b/lib/local/LandmarkDetector/src/LandmarkDetectorParameters.cpp
@@ -148,11 +148,6 @@ FaceModelParameters::FaceModelParameters(vector<string> &arguments)
 			valid[i + 1] = false;
 			i++;
 		}
-		else if (arguments[i].compare("-gaze") == 0)
-		{
-			track_gaze = true;
-			valid[i] = false;
-		}
 		else if (arguments[i].compare("-q") == 0)
 		{
 
@@ -301,7 +296,7 @@ void FaceModelParameters::init()
 	reg_factor = 25.0f;
 	weight_factor = 0.0f; // By default do not use NU-RLMS for videos as it does not work as well for them
 
-	validation_boundary = -0.45f;
+	validation_boundary = 0.725f;
 
 	limit_pose = true;
 	multi_view = false;
@@ -316,7 +311,5 @@ void FaceModelParameters::init()
 	// By default use MTCNN
 	curr_face_detector = MTCNN_DETECTOR;
 
-	// The gaze tracking has to be explicitly initialised
-	track_gaze = false;
 }