Using GPT for Driving Test
This project aims to solve a driving test by using several different machine learning models that extract contextual information from a Dutch driving test.
To use the code, there are some prerequisites
pip install requirements-colab.txt
Next, there are some models that need to be put into the /models folder.
Next, place the images you want to analyze inside the /images folder.
Insert your API key for OpenAI inside the chat.py file.
Create a folder with the name of "Crops" in the working directory
Running main.py will run through all the images in the /images folder. The results are then saved in the /results folder. The results are given in the following format:
- crops
-
- folder including crops detected from all the found objects in the image
- df
-
- folder including the completed dataframe including all information from running the model
- texts
-
- Folder containing all the prompts and responses from each question
- tri-crop
-
- Folder including per question crops of the front view, rear view, and speedometer.
- confusion.png
-
- A confusion matrix based on all the results of the model
- results.csv
-
- Csv file containing the results from all the images