/MTT-Models

Training Various Models for ASL

Primary LanguagePython

Installation (Ubuntu 18.04)

Clone & Install OpenCv

  • git clone https://github.com/primus852/mtt-models
  • cd mtt-models
  • sudo apt install python3-opencv

(optional) Install Virtual Env

  • sudo apt-get install python3-venv
  • python3 -m venv mtt-models-env
  • source mtt-models-env/bin/activate

(optional) Upgrade pip

  • pip3 install --upgrade pip
  • sudo python3 -m pip install -U setuptools

Install dependencies

  • pip3 install -r requirements.txt

Train the Model "LongList" only

  • python3 main.py --long-epochs 2 --skip-short

Train the Model "ShortList" only

  • python3 main.py --short-epochs 10 --skip-long

Run the OpenCV Detection (WIP)

  • python3 main.py --production

Description

This is a standalone Script for 3 things:

  1. Train a long list of Models with a fraction of the Dataset to measure performance
  2. Train a Top X short list with the full dataset and out put a model to feed to OpenCV
  3. Run OpenCV with one of the trained models

First run

The first run may take some time as it will

  • Download the Dataset
  • Extract the downloaded zip
  • Create all needed folders