jijo66/kyc-document-classification

A deep learning model for automating KYC document classification in Kenya # Automated KYC Document Classification ## Overview This project focuses on automating the classification of Know Your Customer (KYC) documents—specifically Kenyan identity cards and passports—using deep learning techniques. The primary goal is to enhance the efficiency of document verification processes in the banking and fintech sectors in Kenya, reducing the reliance on manual verification methods. ## Objectives - Develop a deep learning model to classify KYC documents. - Evaluate the model's performance using various metrics. - Implement a user-friendly web interface for document submission and classification. ## Technologies Used - **Python**: The primary programming language for implementation. - **TensorFlow/Keras**: Libraries used for building and training the deep learning model. - **Streamlit**: A framework for creating the web interface. - **NumPy** and **Pandas**: Libraries for data manipulation and numerical computations. - **Matplotlib**: Used for visualizing training results and performance metrics. ## Dataset The dataset consists of approximately 150 images of Kenyan KYC documents, including both identity cards and passports. The images were sourced from publicly available datasets and are organized into training, validation, and testing subsets. ### Data Preprocessing Before training the model, several preprocessing steps were performed: - Images were resized to a uniform dimension of 180x180 pixels. - Pixel values were normalized to a range of [0, 1]. - Data augmentation techniques were applied to increase dataset variability. ## Model Architecture The model is built using a Sequential Convolutional Neural Network (CNN) architecture, which includes: - Convolutional layers for feature extraction. - Max pooling layers to reduce dimensionality. - Dropout layers to prevent overfitting. - A fully connected layer with a Softmax activation function for multi-class classification. ## How to Run the Project 1. Clone this repository: ```bash git clone …