spook-the-midlands

Facial keypoints of the template
Facial keypoints of the input image
Final result
Delaunay triangulation of the template image
Delaunay triangulation of the input image

Inspiration

Putting make up for Halloween without knowing how it'll end up doesn't sounds good. Trying different Halloween make up without actually putting layers on your skin is time and effort saving. Also it's kinda fun to scare our friends and family.

What it does

The web app asks the user to upload a portrait headshot of the user. Then it'll ask the user to select the scary level(intensity of the makeup). The user then have 3 choices of filter to apply on. And then grab a chair, submit and boom! the result image will scare you.

How we built it

Backend model: We preprocess the photos, to ensure the faces are centre aligned The pre-trained facial landmark detector inside the dlib library computes facial key points Use the landmarks to compute Delaunay triangulation Apply affine transformation between the tuples of triangles from both photos. We generate a mask and impose it on the input image.

Frontend: We used streamlit(a python framework) for creating the frontend We used an input section for uploading the images A slider for choosing the alpha value A drop down menu to select the templates and a output section to display the result

Challenges we ran into

Crop only the faces, and ensure the minimum crop to avoid data loss compute the affine transformation between triangles from both images Apple silicon m1 compatibility issues

Accomplishments that we're proud of

This was our 5th attempt for creating the final result, it was a simple project but we faced a lot of hurdles which took our efforts and time and really checked our patience. But we didn't give up and patiently completed the project in the given time.,

What we learned

Soft: No project is big or small, and every project has its challenges. Handling the pressure of the situation and keeping ourselves composed in such times. Sharing ideas and asking for help is extremely useful.

Technical: Portability is essential. Something might work in some system some might not Create, upload, register and start Azure containers