The Good2Go convenienent stores are paired with gas stations across Utah, Idaho, New Mexico, Colorado, and Arizona. Good2Go is currently interested in trying to evaluate the performance of their stores, one metric they would like to use in this evaluation is the number of lanes that the closest road has to their gas stations. They Reached out and asked the Data Science Society of Brigham Young University-Idaho to create a Convolutional Nueral Network that could predict the amount of lanes on a road based from an image.
This repository contains images from team members that were gathered through API's. In the end we decided to go with 2 metrics to evaluate our model on. The first is accuracy, the second is what we created and decided to call "almost_accuarcy". Almost accuracy allows the lane prediction to be off by 1 lane and will still count that as being accurate.
The final results for our best model are:
(We created our final model from the "stats_gtg_model.ipynb" file)