This project focuses on developing a machine learning model to identify numbers within images, laying the groundwork for future text detection algorithms.
This project aims to classify and localize numerical digits in images, preparing for more complex text detection tasks. It involves preprocessing, training a convolutional neural network (CNN) model using TensorFlow/Keras, and evaluating its performance.
The dataset used consists of images containing single digits. It is divided into training, validation, and test sets with appropriate preprocessing applied.
The CNN model architecture includes convolutional layers, max pooling, dropout for regularization, and fully connected layers with ReLU and softmax activations for classification.
The model is trained using Adam optimizer with categorical crossentropy loss. Data augmentation techniques such as image shifting, zooming, and rotation are applied to enhance robustness.
Model performance is evaluated on a separate test set using accuracy metrics. Training and validation loss/accuracy curves are plotted to assess training progress.
-
Python 3.x: Programming language used for development.
-
TensorFlow 2.x: Open-source machine learning framework by Google.
Build and train deep learning models, manage computational graphs, and optimize GPU computation.
- OpenCV: Library for real-time computer vision tasks.
Image and video processing, including reading, resizing, manipulation, and object detection.
- NumPy: Fundamental package for numerical computing in Python.
Support for multi-dimensional arrays, matrices, and mathematical operations, essential for data manipulation.
- Matplotlib: Plotting library for Python.
Create visualizations such as graphs for training/validation metrics and display sample images.
- scikit-learn: Machine learning library for Python.
Tools for data mining, evaluation of model performance, and potentially preprocessing and feature extraction.
Training: Run python train.py to train the model using the prepared dataset.
Prediction: Run python predict.py to use the trained model for real-time prediction using webcam capture.
Author: Afsheen Wasfiya Abdul Wahab
Dataset: The Chars74K dataset
LinkedIn: https://www.linkedin.com/in/afsheen-wasfiya-abdul-wahab-33a504257