Integrate state of the art facial detection and inverted indexing in the Spark-Tensorflow environment to allow for near real time reverse image search on a roster of faces. Benchmark the accuracy and runtime trades off of state-of-the-art facial detection algorithms and data storage systems for facial detection and inverted indexing. Explore improvements to these algorithms and data storage systems.
|Find archive of video footage that has the faces labeled with bounding boxes.||Week 3 ~ 4|
|Decide on at least two differenet facial recognition systems at different parts of the pareto frontier.||Week 3 ~ 5|
|Set up a Spark-Tensorflow environment. Run a trivial task to get used to the environment.||Week 3 ~ 4|
|Build models||Week 4 ~ 9|
|Benchmark the models and the data storage environments||Week 7 ~ 9|
|Explore efficiency improvements to facial recognition and indexing algorithms||Week 7 ~ 9|
|Explore efficiency improvements to data managment system||Week 7 ~ 9|
|Analize experiments and polish paper||Week 9 ~ 10|
The results of my milestones will go here.