Skip to content

patrick-5546/ML-Image-Processor

Repository files navigation

ML-Image-Processor

The process of organizing photos can be a huge pain. And when they aren't organized, it is still a pain to search through huge unsorted folders, looking for that one photo. This application aims to resolve this dilemma by using machine learning to implement the following features:

Application Features

  1. Category Tags: identifies common objects in an image, using them to tag the image with one or more categories

    • Categories include: Portrait, Group Photo, Urban, Pet, Nature, Sports, Food
    • The objects that the models are trained to detect, and how they map to the categories, are listed in tagger/categories.py
  2. Name Tags: identifies people in an image and tags the image with their names

  3. Rotation Correction: detects crooked images and straightens them

  4. Low-Light Enhancement: detects dark images and brightens them

    • Beta feature: may output images with bright neon spots

Video Presentation

Video presentation

Installation

  1. Ensure that Python>=3.9 (with pip) is installed

  2. Download the application

    • To get only the files required to run the application, download and unzip a Release
    • To also get the reports, datasets, notebooks, etc., clone the repository (much larger folder):
  3. Create a virtual environment to run the application in (optional, but recommended)

    1. Navigate to the project directory
    2. See Python Docs
      • For the first command, use venv as the directory name instead of tutorial-env
      • If using PowerShell, run the Activate.ps1 script instead
    • This ensures that the packages necessary to run the application are used without adding to or replacing packages in the main python package directory
  4. In the project directory, install the required packages

    pip install -r requirements.txt

How to Run

  1. Run python main.py to start application
  2. Access application at http://127.0.0.1:8080/
  3. Application usage instructions can be found on the homepage, and a visual walkthrough in the final report