JMoonr / LATR

[ICCV2023 Oral] LATR: 3D Lane Detection from Monocular Images with Transformer

https://arxiv.org/abs/2308.04583

MIT License

184 stars 31 forks source link

Open AZKvgkv opened 1 week ago

AZKvgkv commented 1 week ago

Data Preparation

Follow OpenLane to download dataset and then link it under data directory.

cd data && makir openlane && cd openlane
ln -s ${OPENLANE_PATH}/images .
ln -s ${OPENLANE_PATH}/lane3d_1000 .

The first instruction should be

cd data && mkdir openlane && cd openlane

usst-st commented 1 day ago

LATR/docs /train_eval.md

Apollo
- Visual Variations CUDA_VISIBLE_DEVICES='0,1,2,3' python -m torch.distributed.launch --master_port=29284 --nproc_per_node 4 main.py --config config/release_iccv/apollo_standard.py --cfg-options evaluate=true eval_ckpt=pretrained_models/apollo_illu.pth should be CUDA_VISIBLE_DEVICES='0,1,2,3' python -m torch.distributed.launch --master_port=29284 --nproc_per_node 4 main.py --config config/release_iccv/apollo_illu.py --cfg-options evaluate=true eval_ckpt=pretrained_models/apollo_illu.pth