A simplified and improved academic project implementing a CNN from scratch using NumPy to detect driver distractions. D3-CNN-2 achieves ~86% test accuracy in just 11 epochs, outperforming the original model with fewer layers. Includes a Streamlit-based frontend for real-time image/video testing — all without using PyTorch or TensorFlow.
machine-learning computer-vision deep-learning neural-network cnn driver-distraction-detection semeter
-
Updated
May 21, 2025 - Jupyter Notebook