Feature detection

[seanavery]

Problem

Using every pixel as a feature say 50 px * 50 px = 2500 features... way to much

Solution

not sure yet.. edge detection? Break down into smaller feature set: jaw, eyes, ears, hairline, ...

[seanavery]

1. Bound the face in the picture
2. Detect key facial features
   • [x] eyes
   • [x] nose
   • [x] mouth
   • [x] eyebrows
   • [x] jaw

get coordinates for all of these features using dlib's python api

[seanavery]

Problem

dlib is breaking on mach-o

built for simulator not macOS

[seanavery]

still having issue with mach-o...

solution

• [x] fork and integrate as submodule
• [x] build from source (compiling c++ code, ...)

[seanavery]

opencv mean-shift to target a box around face

[seanavery]

face and eye detection via opencv's haar cascade pre-trained models

[seanavery]

dlib finally running and feature detection working.

python3 ./scripts/detect_features first_name last_name