Home

Bikeway Obstacle Detection using Smartphone

Bikeway Obstacle Detection using Smartphone (B.O.D.S.) is a system that uses newer smartphones to capture images and depth data from the environment. These images are processed to create a model that predicts if an obstacle is near or approaching the rider's path. The current version of B.O.D.S. utilizes an iPhone 13 Pro and a binary classification CNN model with F1 score as the performance metric. Still, other smartphones and machine learning techniques and models can be explored. The document provides a baseline for contributors and highlights areas for system manipulation. Further work is needed to collect more types of ground obstacles and experiment with different models.

Data Collection with ObstaSpy

1. Installation guidelines and minimum requirements
2. Pre-capture Camera Alignment and Steam Settings
3. Floating-point Depth Data to Grayscale PNG
4. Image Output File Settings

Obstacle Detection Model Development

1. Installation guidelines and minimum requirements
2. Image Conversion Process for Training
3. Obstacle Detection Algorithm for Binary Classification
4. Application of Adaptive Thresholding and Morphological operations
5. Find the contours of an obstacle in a region of interest (ROI)
6. Summary of Model Architecture
7. Summary of Performance using F1 Score