Object Detection Model:

Release Notes version Firearm Detection 1.0

NEW FEATURES

Two different running modes
More accurate and efficient model. Version 6 of the model

BUG FIXES

Handle keyboard interrupt to exit gracefully in the middle of a video or during live feed.

KNOWN BUGS

Local video mode will not exit gracefully at the end of the video. 
For an unkown reason opencv believes there is one additional frame.
Command line does not print if a gun is detected. Could not find a way to access confidence scores.

System Requirements

Install instructions are for Linux and tested specifically on Ubuntu 16.04.
Install script is made for a computer using a 64 bit architecture in the stated enviornment.

Install Instructions

Run the following commands to install git, clone our repository, and run our installation script

$sudo apt-get install git
$git clone https://github.com/GT-FiniteLoop/object-dectection-model.git

$cd object-detction-model
$chmod +x ./tf_install.sh

$sudo ./tf_install.sh

The install script will install python 3.5, pip for python 3, tensorflow 1.13, tensorflow dependencies, and protoc.
It will clone the tensorflow models repository as well, which is used by our application.
Protoc is used by tensorlow to compile some C++ into python. 

Run Instructions

The two implemented modes are live and local.
Live will take the first available camera and local will take a provided video file.

python3 model.py  --mode live
or
python3 model.py --mode local --video <video file>