GitHub - joonjeon/PINet_new: This repo attempts to convert PINet to TensorRT

Update

2020-11-06: new trained model ("0_tensor(0.5242)_lane_detection_network") (Acc: 96.81%, FP: 0.0387, FN: 0.0245, threshold = 0.36)

Pre-trained model is "804_tensor(0.5786)_lane_detection_network.pkl"

You can download the dataset from https://github.com/xbjxh/CurveLanes. Pre-trained model is "savefile/32_tensor(1.1001)_lane_detection_network.pkl"

Here are several scripts for parsing and converting datasets to TuSimple style.https://github.com/pandamax/Parse_Curvelanes

Name		Name	Last commit message	Last commit date
Latest commit History 84 Commits
CULane		CULane
CurveLanes		CurveLanes
TuSimple		TuSimple
LICENSE		LICENSE
README.md		README.md
agent.py		agent.py
data_loader.py		data_loader.py
evaluation.py		evaluation.py
export.py		export.py
fix_dataset.py		fix_dataset.py
hard_sampling.py		hard_sampling.py
hourglass_network.py		hourglass_network.py
onnx_converter.py		onnx_converter.py
onnx_inference.py		onnx_inference.py
parameters.py		parameters.py
pinet.engine		pinet.engine
pinet.onnx		pinet.onnx
test.py		test.py
test_label.json		test_label.json
test_quantized.py		test_quantized.py
test_result.json		test_result.json
train.py		train.py
util.py		util.py
util_hourglass.py		util_hourglass.py