James Menden, Sean Felderman, John Bolka

CS 534: Fall 2018 - Professor Mohit Gupta

Project Proposal

Idea:

Our idea for the final project is to use face detection from a collection of images to find all faces in these files. Once all of the faces are found and isolated, there will be a single face (either from one image we select or a randomly selected face from the faces we already processed if one is not provided) that will serve as a baseline. We will use components from all the faces that we processed originally in the other files to create a joiner. The final image that the joiner will make will be a picture of the goal face from the input file that was chosen. This process is also known as photographic mosaic, which is a very popular way of creating compelling, engaging and thought-provoking artwork. As far as image processing techniques, we will need to use edge detection to find the edge of the face and where to stop cropping for our face, image alignment to rotate the face to the standard alignment of the goal image.

As a stretch goal, we want to give the option to choose how choppy or smooth the image is. Where users can pick ‘choppy’ where the image looks like a Picasso painting, or ‘smooth’ where it looks closer to a normal human face.

Timeline:

Sep 26th - Oct 10th: Research

• Determine image collection- either through large collections of online images or a preselected groups of faces
• Read applicable publications on face detection and stitching together of images

Oct 8th: Image Alignment and Panorama Creation Lecture

Oct 10th: Face Detection Lecture

Oct 10th - 31: In this period we aim to have the image selection process worked out.

• Images will be cropped to include only the relevant portions including faces
• Images will also be rotated/aligned so that they are in a consistent format before we attempt to collage them

Oct 31: Midterm Report Due

Oct 31 - Late Nov: Joining images

• The second stage of this project is “joining” the images together to form one singular face.
• This will require taking our base image and fitting together regions of similarity to eventually form the completed image

Stretch Goals:

• Should we have additional time, we will work on creating a smoother collage, see stretch goals above
• This may require finer sampling of the edge regions to find images that more closely match to provide a continuous experience.

Dec 3-12: Presentation

Dec 12: Website Due