/poker-card-image-recognition

This is an image recognition project, in which we will be classifing a set of poker card into 53 categorises of card types,and training a model to classify image of cards into any of the 53 categorises.

Primary LanguageJupyter Notebook

Poker Card Image Recognition

This is a machine learning project that uses convolutional neural networks (CNN) to classify poker cards based on their type. The goal of the project is to develop a model that can accurately identify a given card image as one of 53 different categories of cards.

Dataset

The dataset used for this project is the Cards Image Dataset-Classification, which contains over 8000 images of poker cards in JPG format. The dataset is divided into three sets: training (7624 images), validation (265 images), and testing (265 images). The images are of size 224 x 224 pixels with three color channels.

Methodology

The project follows a standard machine learning workflow, including the following steps:

  1. Data Collection: The dataset was downloaded from Kaggle and examined to gain insights into the data.

  2. Data Preprocessing: The images were preprocessed by resizing them to a standard size and converting them into an array format suitable for feeding into our CNN model.

  3. Data Augmentation: Data augmentation techniques were used to increase the size of the dataset and improve the robustness of the model. Techniques used include image rotation, flipping, and zooming.

  4. Model Building: A CNN model was built using Keras with TensorFlow backend, and trained using the preprocessed dataset. Different architectures, hyperparameters, and optimization algorithms were experimented with to achieve optimal performance.

  5. Model Evaluation: The performance of the model was evaluated using various metrics such as accuracy, precision, recall, and F1 score. The results were visualized using a confusion matrix and classification report.

  6. Model Deployment: The trained model was deployed to make predictions on new, unseen poker card images. The predictions were visualized.

Getting Started

To run this project, you will need to have the following tools and libraries installed:

  • Python 3.8+
  • Keras
  • TensorFlow
  • NumPy
  • Matplotlib
  • Scikit-learn

You can run the project on your local machine or in a Jupyter notebook environment such as Google Colab.

  1. Clone this repository to your local machine:
git clone https://github.com/yourusername/poker-card-image-recognition.git

  1. Download the dataset from Kaggle and extract it to the data folder in the project directory.

  2. Open the Jupyter notebook poker-card-image-recognition.ipynb and run the cells in order.

  3. Once the model is trained, you can use it to make predictions on new poker card images.

Credits