Implement new facial detection/recognition engine

[ryspark]

Current model is vanilla FaceNet + MTCNN. Possible enhancement include:

FaceNet enhancements (facial recognition):

• Adaptive threshold: https://arxiv.org/pdf/1810.11160.pdf

FaceNet replacements (facial recognition):

• https://github.com/wang-xinyu/tensorrtx/tree/master/arcface, https://arxiv.org/pdf/1801.07698.pdf
• https://github.com/deepinsight/insightface

MTCNN replacements (facial detection):

• https://github.com/deepinsight/insightface/tree/master/detection/retinaface
• https://github.com/deepinsight/insightface/tree/master/detection/scrfd

[ryspark]

WEEK 3 GOALS

Andrews + Riya: adaptive threshold (2 more weeks?)
Alex, Andy, Jason: set up Google Cloud, get https://github.com/wang-xinyu/tensorrtx/tree/master/arcface code running maybe?

[ryspark]

WEEK 5 GOALS

Everyone:

1. Read up on arcface - see paper https://arxiv.org/pdf/1801.07698.pdf - jason, riya, andrew 2x, andy
2. Try to install tensorrt on Windows 10 (including cuda, cudnn) - jason, andy, andrew h

[ryspark]

WEEK 6 GOALS

• Andy, Ria, Andrew H, Jason: get Google Cloud VM set up

[ryspark]

WEEK 7 GOALS

• Everyone:

1. Finish updating Google Cloud VM
2. Install pycuda and tensorrt.
3. Clone aisecurity repository

[ryspark]

WEEK 8

1. Finish installing VM on Google Cloud

• After booting it up, select y when it asks to install drivers
• Then, follow the instructions here (first answer): https://askubuntu.com/questions/1109982/e-could-not-get-lock-var-lib-dpkg-lock-frontend-open-11-resource-temporari
• Then, run sudo /opt/deeplearning/install-driver.sh

2. Install pycuda and tensorrt.

[EnergeticHay]

WEEK 9

1. Finish installing pycuda and tensorrt on your VM
2. Clone aisecurity, configure it
3. Use opencv to open up the video and loop through it (use the aisecurity API)

[novialriptide]

It looks like PR #31 closed this automatically, reopened.

[ryspark]

WEEK 10

1. Continue VM installation process
2. Clone aisecurity on your local computers and test it