CaliFree3DLane

• 🚘 The complete code will be released after work is accepted

Environment Setup

• We develop with PyTorch 1.8 and recommend you to use Anaconda to create a conda environment before installing the dependencies

pip install torch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0

• For Deformable Attention

cd models/ops/
bash make.sh
# unit test (should see all checking is True)
python test.py

• For other dependencies

pip install -r requirement.txt

Training and evaluation on Apollo 3D Lane Synthetic

• How to train:
  1. Please download Apollo 3D Lane Synthetic and modify the configuration in the /tools/apollo_config.py
  2. Execute the following code:

cd tools
python3 train_apollo.py

• How to evaluate:
  1. Please modify the configuration in the /tools/val_apollo.py
  2. Execute the following code:

python val_apollo.py

Training and evaluation on OpenLane

• How to train:
  1. Please download OpenLane and modify the configuration in the /tools/apollo_config.py
  2. Execute the following code:

cd tools
python3 train_openlane.py

• How to evaluate:
  1. Please modify the configuration in the /tools/val_apollo.py
  2. Execute the following code:

python val_openlane.py

Benchmark

Results of different models on Apollo 3D Lane Synthetic (Balanced Scence)

• In/Extrinsics

Method	F-Score	X error near	X error far	Z error near	Z error far
3D-LaneNet	86.4	0.068	0.477	0.015	0.202
Gen-LaneNet	88.1	0.061	0.486	0.012	0.214
CLGO	91.9	0.061	0.361	0.029	0.25
PersFormer	92.9	0.054	0.356	0.010	0.234
Anchor3DLane	95.4	0.045	0.300	0.016	0.223
BEV-LaneDet	98.7	0.016	0.242	0.020	0.216
Ours	97.9	0.027	0.248	0.021	0.221

• without In/Extrinsics

Method	F-Score	X error near	X error far	Z error near	Z error far
BEV-LaneDet	96.8	0.035	0.292	0.033	0.250
Ours	97.9	0.031	0.261	0.030	0.243

Results of different models on OpenLane dataset

• In/Extrinsics

Method	F-Score	X error near	X error far	Z error near	Z error far
3D-LaneNet	44.1	0.479	0.572	0.367	0.443
Gen-LaneNet	32.5	0.591	0.684	0.411	0.521
PersFormer	50.5	0.485	0.553	0.364	0.431
Anchor3DLane	54.3	0.275	0.310	0.105	0.135
BEV-LaneDet	58.4	0.309	0.659	0.244	0.631
Ours	58.8	0.273	0.687	0.201	0.626

Method	All	Up& Down	Curve	Extreme Weather	Night	Intersection	Merge& Split
PersFormer	42.0	40.7	46.3	43.7	36.1	28.9	41.2
Anchor3DLane	54.3	47.2	58.0	52.7	48.7	45.8	51.7
BEV-LaneDet	58.4	48.7	63.1	53.4	53.4	50.3	53.7
Ours	58.8	51.8	62.9	56.4	54.2	50.9	53.5

• without In/Extrinsics

Method	F-Score	X error near	X error far	Z error near	Z error far
BEV-LaneDet	54.7	0.346	0.769	0.253	0.709
Ours	57.0	0.306	0.761	0.227	0.708

Method	All	Up& Down	Curve	Extreme Weather	Night	Intersection	Merge& Split
BEV-LaneDet	54.7	47.4	61.4	49.2	49.7	46.4	49.9
Ours	57.0	48.9	62.3	54.8	52.0	50.2	52.1

Visualization