Setting Up an MVP

January 27, 2019

Crunch time is coming up! Out first big deadline is coming up soon - on February 6th, we're demoing our product to the professors. For our basic demo, we'd like to get a two big things working. The first of these is to 3d print and assemble our mechanical mount.

Exploded CAD View

The second thing we'd like to showcase is our face detection and tracking algorithm. We've already successfully run person detection algorithms by combining OpenCV with YOLO net's object detection. We've also tested several tracking algorithms, including KCF, BOOSTING, MIL, GOTURN, etc. Currently, KCF (kernalized correlation filters) seems to offer the best blend between performance and accuracy, since several of these algorithsm are either fast but inaccurate (stop tracking correctly) or slow but accurate (unable to run on realtime footage).

Object Detection with OpenCV
For our remaining goals, we'd like to finish up with the two above stated goals and get a functional demo for the profs. We'll be focusing on merging the tracking and detection algorithms to correctly identify people coming in and out of a door. Stay tuned for a happy February 6th update!