rv2442 / gtsrb_mnist_model Goto Github PK

Add comments on for all .py files explaining work flow and logic

Create a model for Lane Detection and Steering Prediction.

Goal: Current Lane must be detected and a prediction of car's position with respect to the center of the lane must be given as output or by how distance does the car need to be Steered back to the center of the lane must be given as output.

Steps:

1. Collect Data from custom made Track
2. Sample and Segregate Data
3. Preprocess (Image Processing and Normalisation)
4. Design NN architecture (Tensorflow)
5. Train, Validate and Test
6. Predict Single Image
7. Predict Live video
8. Try for Higher accuracy and optimize
9. Implement on Jetson Nano and Tweak accordingly.

Result: A model working on Jetson Nano that can be recognised to work as a subset of an ADAS feature of Lane Steering ( maintain itself in the current lane).