Realtime self driving car implementing DNN

Install dependencies and create environment

First install the Anaconda distribution and install all the required dependencies.Create the virtual environment tf-gpu (for eg)

conda create --name tf-gpu

install supported verson of python by tensorflow and proceed to below steps

Run the pretrained model(model-010.h5)

Turn on all the connections to esp32 and H-bridge driver and wait for five seconds while the servo returns to the centre

python drive.py model-010.h5

To train the model

You'll need the data folder which contains the training images.

The folder is created automatically when you run

python train.py

This creates a folder named training_data and run below code corresponding to the created folder!

python model.py

if you run python train.py more than once the code automatically creates training_data1 and so on!

After running model.py it will generate a file model-<epoch>.h5 whenever the performance in the epoch is better than the previous best. For example, the first epoch will generate a file called model-000.h5.Models are saved after every epoch so run the latest model when you run the drive.py

Also,plese checkout the video by clicking on the thumbnail below: