Convolutional Neural Network (CNN) is a deep learning algorithm that learns directly from data, eliminating the need for manual feature extraction. CNNs are particularly useful for image data, helping recognize patterns in images. In this project, we build a CNN model for image classification, categorizing images into classes such as social security cards, driving licenses, and others. We have used PyTorch for building the model, which offers dynamic computational graphs and a Pythonic interface.
To build a Convolutional Neural Network model to classify images into different classes in PyTorch
- Language:
Python - Libraries:
PyTorch,pandas,matplotlib,NumPy,opencv_python_headless,torchvision
The dataset includes images of driving licenses, social security cards, and other categories. These images are of varying shapes and sizes, which are preprocessed before modeling.
- Data Loading
- Data Preprocessing
- Resizing and scaling of the images
- Encoding of the class labels
- Model Building and Training
- CNN model building in PyTorch
After unzipping the pytorch_cnn.zip file, you'll find the following folders:
- Input: Contains training and testing data for image classification.
- Notebook: Contains the Jupyter notebook file for the project.
- ML_Pipeline: A folder with Python functions organized in different files. These functions are called inside
Engine.py. - Output: Contains the saved CNN model.
- requirements.txt: Lists all the required libraries with respective versions. Install them using
pip install -r requirements.txt. Readme.md: This file with instructions for running the code.
There are two ways to execute the end to end flow.
- Modular Code
- IPython
- Create virtualenv
- Install requirements
pip install -r requirements.txt - Run Code
python Engine.py - Check output for all the visualization