A neural network that performs object detection on images containing pedestrian traffic lights to help visually impaired pedestrians navigate safely outdoors.
The neural network was trained using TensorFlow's Object Detection Library.
The file detection-inference-test.ipynb can be used to run object detection inference (recommended to run using Google Colab).
The file Copy_of_tensorflow_object_detection_training_colab.ipynb was used to train the neural network (Also recommended to run using Google Colab).
The Utils folder contains python files used to load and process the training and validation images used.
The Models folder contains saved TensorFlow model files.
The Python library was used for the majority of the neural network training.
Python libraries used include TensorFlow, The XML Element Tree Library, and The OpenCV library
Images used for training and validation were obtained from google images.
The LabelImg software for annotating and labeling images
The neural network's output appears similarly to the following images:
More information about this project, including background information, results, and interpretation of results can be found in the Project Thesis file