CrosswalkVisionPython propose a PyTorch algorithm to predict the position of the entry and the exit of crosswalk. This project is inspired by the project ImVisible also available on github. This project was created for a CentraleSupelec classes.
The file requirements.txt list all the package necessary to run the project. Also you will have to download all the dataset and put it inside the dataset folder which is currently empty.
You can download the dataset with those two links: Part 1 and Part 2 (23Go).
The file Model/train.py is used to train the model, and the file Model/test.py is used to check the perfomance of the train model.
The file Model/NeuralNetwork.py encode the model in PyTorch.
The folder TrainingModel save the model at different epoch and also some charts.
Finally, the file CoreML.py is used to convert the PyTorch model into a CoreML model.