Deployed through Heroku: https://epic-flask3.herokuapp.com/
A three-tiered demonstration of machine learning applied to facial images:
(in order of priority and increasing difficulty)
Tier 1 – Base, “must do”: Identify key facial points of a face image. Using the points, add accessories to the face as a demonstration of application.
- dataset: https://www.kaggle.com/c/facial-keypoints-detection/data
- Show input image
- Show image with points identified
- Show image with accessories aligned to points
Tier 2 – Reach, “will try to do”: Determine gender and estimate age of subject from a facial image.
- dataset: https://software.intel.com/en-us/articles/efficient-computation-on-the-edge-with-intel-movidius-neural-compute-stick?utm_source=ISTV&utm_medium=Video&utm_campaign=ISTV2018_ISTV1815
- “This image is a 37 year old female”
- “This image is a 20 year old male”
Tier 3 – Stretch, “if the stars align”: Generate an image based on text description of a face:
- dataset: https://software.intel.com/en-us/articles/efficient-computation-on-the-edge-with-intel-movidius-neural-compute-stick?utm_source=ISTV&utm_medium=Video&utm_campaign=ISTV2018_ISTV1815
- “30 year old Asian male”
- “50 year old Caucasian female”
All of these have more than one example with code available. Tier 1 has trained models that can serve as a starting point to work off of.
We will demonstrate these with “canned” demos showing a sample image, the result of the processing and transformation. As a stretch, if all is going well, we may have live demonstration. We will not be wrapping these in applications to allow us to dedicate the most time to machine learning functions.