Using OpenCV to detect and extract faces
My thesis will be utilizing different tools to sort and filter images. Among them, face detection and face recognition. OpenCV has a robust set of optimizations for face and facial feature detection based on the Viola-Jones object detection framework. Working with Python master, Akira Shibata, we quickly had promising results:
Shown here are results with optimizations for Frontal Face, Eyes, Nose, and Mouth:
Next we integrated the OpenCV detection with some python code from Bytefish’s blog and were able to successfully extract images from a folder of multiple photos. The accuracy was heavily determined by resolution which makes sense, though I was surprised at times by how fuzzy or blurred the regions were where faces were successfully extracted.
Shown here is an initial face extraction of 40 faces from multiple photographs (repeats are due to multiple images at different resolutions):
The next challenge is to begin face recognition using a Fisher Faces model with the extracted faces. A literature review on the core topic of Human-Assisted image clustering that continues to grow.