The purpose of this project is to train a lane line segmentation model for an advanced driver-assistance system. More about this project. The inference code was integrated in this repository.
Create anaconda enviroment:
conda create --name <ENVIROMENT_NAME> python=3.6
Activate created environment and install all requirements:
pip install requirements.txt
Create new config file in list_config
directory Please don't modify old config file, so we can have better observation, model and training history will be auto saved into saved_models
folder.
For training, simply run:
python model/train.py
or
./train.sh
-
Modify model paths in
convert_h5_to_pb.py
andconvert_pb_to_uff.py
. -
Convert
.h5
model to.pb
, and finnally.uff
:
pip install requirements-h5-to-uff.txt
python convert_h5_to_pb.py
python convert_pb_to_uff.py