How to 

1.	Recognize_points.py in the Preparation folder reads in the images that are in FRS/with_points and in FRS/without_points. The program automatically recognises the marked reference points and saves all coordinates and the original image in the folders 1_Coordinates_For_Image and 1_Images_Without.

2.	The training and testing procedures are done for each point individually, so there is a separate folder for each cephalometric landmark. Each folder contains a script extract_point.py extracting the specific point from all coordinates (this requires a complex algorithm and must be done differently for each point). The coordinates of this point are then saved in ../../CNN/S_Point/Coordinates. The images with and without points are also saved in ../../CNN/S_Point/Images_with and ../../CNN/S_Point/Images_without.

3.	The CNN folder contains all the files needed to train and test the model. Train_CNN.py is used for training and the trained CNN is saved for each point in the corresponding folder.

4.	Recognize_Image.py uses the trained model to recognise the cephalometric points.

5.	Evaluate_Recognize_Image.py applies the trained neural network to detect the cephalometric points on the images and then stores the mean radial error, the error along the X-axis (∆x) and correspondingly the error along the Y-axis (∆y) in txt files located in ../../CNN/Test_Data/External_Validation

6.	In order to run Evaluate_Recognize_Image.py, the folders for each point in ../../CNN/Test_Data/External_Validation/S_Point_Coordinates and .../External_Validation/S_Point_images_without must be prepared manually.