【实时建图】MapTR（4）------训练

1 下载预训练权重

• backbone的预训练权重

mkdir ckpts
cd ckpts
wget https://download.pytorch.org/models/resnet50-19c8e357.pth   # resnet backbone
wget https://download.pytorch.org/models/resnet18-f37072fd.pth

• maptr 权重

2 设置训练参数

修改/root/MapTR/projects/configs/maptr/maptr_nano_r18_110e.py

# 1. 预训练权重
pretrained=dict(img='ckpts/resnet18-f37072fd.pth'),# 2. BN层修改 单卡BN，多卡用SyncBN
# norm_cfg=dict(type='SyncBN', requires_grad=True), # 多卡BN层
norm_cfg=dict(type='BN', requires_grad=True),    # 单卡BN层# 3. 数据集位置
data_root = '/root/nuScenes/'# 4. samples_per_gpu和workers_per_gpu
samples_per_gpu=6,
workers_per_gpu=2,# 训练周期
total_epochs = 25

3 训练

python MapTR-main/tools/train.py MapTR-main/MapTR-main/projects/configs/maptr/maptr_nano_r18_110e.py  --gpus 1 --no-validate