Image Classifier

Project description

Image Classifier is a deep learning-based project designed to classify images into predefined categories. This project uses Convolutional Neural Networks (CNNs) to identify and classify images based on their visual features. The classifier is implemented in Python, utilizing the PyTorch library for building and training the deep learning model.

With Image Classifier is possible to:

• Use existing models to classify images
• Train a new model, simply knowing the categories of the model

Installation

1. Clone the repository:

   git clone https://github.com/SalerSimo/Image-Classifier.git
   cd Image-Classifier

2. Install the required Python packages:

   pip install -r requirements.txt

Usage

1. Open file "launch.cmd" or write in the command line:

   python src/image_classifier.py

2. Select if use an existing model or train a new model:

   • Use model:
     • Select the model
   • Train new model:

     • Insert categories

     • Insert number of images to download for each category

       The new model will be saved into models folder

   All the models must have the following name format:

    CATEGORY-1_CATEGORY-2_..._CATEGORY-N.pth
   

   where CATEGORY-1 ... CATEGORY-N are the categories the image will be classified into.

3. Select the image to classify.

TEST

• Train a new model:

  The first test is made with DOG and CAT. I trained a new model just by inserting Dog and Cat as categories: I used this image as a test:

  

  The result is: