My main aim is to remove the background from facial images that people sent me of their faces. People send me their pictures (.jpg / .png) and I want to convert them into face masks and I want to print them on A4 using the following dimensions: 2480 x 3508 pixels. – I only want to print the face and hair (head hair/beard). I want the face and eyes automatically recognized because of that we can easily extract the face and the background should be removed around the head (and also any of the body that is not the face). I assume that the background removal can be done using any machine learning library or any other accurate face detection method. Facial landmark detection method is the best to detect the facial features like eyes, lips, and nose. Once the background is removed, it should automatically smooth the edges around the face.
I wish for there to be an interface similar to this:
[login to view URL]
The user will be able to add and remove elements of the picture/background and manually change the position of the image to fit on certain eye hole templates. Certain functions such as angling the image and resizing it easily need to be accessible.
The eyes need to be recognized so the image can be automatically placed over one of 8 different options of present "eye hole" areas. The eye holes need to be a certain distance apart so when you wear the mask - you can see out of it. Here is an example of the eye hole template (please note that the number 1 eye hole is where the pupils currently are):
[login to view URL]
If the face does not fit into these eye hole areas automatically due to the head not fitting within the print dimensions - it needs to automatically size the face to eye hole 1 so it's the correct distance apart and prompt the user to place the image in the best position while categorizing it as an “EHU” output file name (see below).
Please read this below paragraph for the exact understanding of EHU:
Depending on which eye hole the face matches up to, will determine the output file name given to it. For example - if it fitted to eye hole 1 - the output name would be EH1_XAMOUNT_USERNAME or eyehole 2 - EH2_X10_USERNAME - if it doesn't fit any of the eyeholes as discussed above the name will be EHU_X200_USERNAME
I would like all files that are able to match to a filename to output in 2 file types – PNG and PDF (multiples will only be applied to PDF – see later).
Each image will also require the file name to be imposed at the bottom of the image.
There will also be a second file produced for each image - which is a backing template. There are 9 different backing templates depending on which eye hole the image matches up to. The filename should also be on this backing template also.
The software should automatically have a “hot folder” assigned to it.
I have attached input image and output image see that both images.