What: Face recognition.
Why: The scope of this project is: (a) to understand the computational challenges of a very popular biometric attribute, i.e., face; (b) to be able to search for your own sources and have a literature review; (c) to construct your own solution, possibly using existing methods, from your own definition of the problem.
How: This will be a biometric validation task.
(a) Collect a set of face images. Best way to do this is to find publicly available datasets. The number of subjects/classes is open (any), and the number of images per subject is also open, however, there should be more than 20 subjects with more than 2 images each.
(b) Find and use any method suitable for face recognition. You may use older methods or more recent deep Neural Network-based frameworks. For the latter you can use models pre-trained on the specific dataset you use.
(c) Pick any three images from the dataset, each from a different subject, as “unknown” query and show the recognition performance. Also, pick two subject images not in the dataset and show the validation result.
Note: any detail not mentioned above is open to your interpretation.
SUBMIT:
1. Matlab/Python/C++ code with any additional external functions you used. [70pts]
2. A screen recording of your execution, also showing some content of the dataset. [20pts]
3. A readme.txt that includes all sources you used, papers, code, data. [10pts]