This project is regarding the stanford-cars-dataset which is available on kaggle. Scope of this project is to predict the car size of a given car image using a convolutional neuronal netowrk (CNN).
- Dean Tomanelli
- Christopher Wolf
open a terminal inside the project folder. While running docker desktop in background enter the following cmd:
docker build -t tf_env .
after the image was successfully built, you can run the container with
docker run -p 8888:8888 --name car_body_prediction tf_env
To test the Model, you have to run the imports cell. For loading the model, scroll down to the "Load Model" cell.
NOTE: the container provides all dependencies to load the final model (you can ignore the warnings and erros). because there are some dependencies which require an nvidia GPU, you will not be able to run the entire model notebook inside the container (i.e. train the model).
running the project outsite the provided container requires an python environment with all dependencies mentioned in requirements.txt again, for running the model.ipynb notebook completly you will need to have an nvidia GPU or adjust the code for CPU usage. Because computational benefit, we've used the GPU for the Notebook.
For using GPU it's highly recommended to use a Anaconda virtual enviornment. With that it's easy and fast to setup all dependencies for the GPU usage.
click here for more details setting up tensorflow with GPU usage.