BEVFusion（MIT）

NUNA、

已于 2023-12-27 11:11:33 修改

阅读量948

点赞数 10

文章标签： linux 运维服务器

于 2023-12-26 14:29:12 首次发布

本文链接：https://blog.csdn.net/weixin_43537097/article/details/135160182

版权

服务器平台: autodl

cuda: 11.1

torch: 10.1

python: 3.8.0

使用nuScenes-mini

（1）首先拷贝文件，再解压

cp /root/autodl-pub/nuScenes/Fulldatasetv1.0/Mini/v1.0-mini.tgz  /root/autodl-tmp

tar zxvf /root/autodl-tmp/v1.0-mini.tgz -C /root/autodl-tmp/bevfusion/data

#正常版本，结构如下：
bevfusion-mit
├── tools
├── configs
├── data
│   ├── nuscenes
│   │   ├── maps
│   │   ├── samples
│   │   ├── sweeps
│   │   ├── lidarseg (optional)
│   │   ├── v1.0-test
|   |   ├── v1.0-trainval
 
 
#如果下载的是mini版本，结构如下：
bevfusion-mit
├── tools
├── configs
├── data
│   ├── nuscenes
│   │   ├── maps
│   │   ├── samples
│   │   ├── sweeps
│   │   ├── v1.0-mini

（2）执行结果如下：

（3）运行数据转换的脚本(注意该命令需要再bevfusion的目录下执行)

#同样的，如果是正常版本，运行：
python tools/create_data.py nuscenes --root-path ./data/nuscenes --out-dir ./data/nuscenes --extra-tag nuscenes
 
#如果是mini版本，运行：
python tools/create_data.py  nuscenes --root-path ./data/nuscenes/ --version v1.0-mini --out-dir data/nuscenes/ --extra-tag nuscenes

结果如下

#正常版本：
├── data
│   ├── nuscenes
│   │   ├── maps
│   │   ├── samples
│   │   ├── sweeps
│   │   ├── lidarseg (optional)
│   │   ├── v1.0-test
|   |   ├── v1.0-trainval
│   │   ├── nuscenes_database
│   │   ├── nuscenes_infos_train.pkl
│   │   ├── nuscenes_infos_val.pkl
│   │   ├── nuscenes_infos_test.pkl
│   │   ├── nuscenes_dbinfos_train.pkl
 
#mini版本
├── data
│   ├── nuscenes
│   │   ├── maps
│   │   ├── samples
│   │   ├── sweeps
│   │   ├── v1.0-mini
│   │   ├── nuscenes_database
│   │   ├── nuscenes_infos_train.pkl
│   │   ├── nuscenes_infos_val.pkl
│   │   ├── nuscenes_dbinfos_train.pkl

下载预训练权重

./tools/download_pretrained.sh

结果：

执行该命令可能会出现

./tools/download_pretrained.sh



--2023-12-26 10:24:00--  https://bevfusion.mit.edu/files/pretrained_updated/bevfusion-det.pth
Resolving bevfusion.mit.edu (bevfusion.mit.edu)... 18.25.16.171
Connecting to bevfusion.mit.edu (bevfusion.mit.edu)|18.25.16.171|:443... connected.
HTTP request sent, awaiting response... 301 Moved Permanently
Location: http://hanlab.mit.edu/projects/bevfusion/files/pretrained_updated/bevfusion-det.pth [following]
--2023-12-26 10:24:01--  http://hanlab.mit.edu/projects/bevfusion/files/pretrained_updated/bevfusion-det.pth
Resolving hanlab.mit.edu (hanlab.mit.edu)... 52.197.0.54, 54.178.223.218, 52.199.221.217
Connecting to hanlab.mit.edu (hanlab.mit.edu)|52.197.0.54|:80... connected.
HTTP request sent, awaiting response... 301 Moved Permanently
Location: https://hanlab.mit.edu/projects/bevfusion/files/pretrained_updated/bevfusion-det.pth [following]
--2023-12-26 10:24:02--  https://hanlab.mit.edu/projects/bevfusion/files/pretrained_updated/bevfusion-det.pth
Connecting to hanlab.mit.edu (hanlab.mit.edu)|52.197.0.54|:443... connected.
HTTP request sent, awaiting response... 404 Not Found
2023-12-26 10:24:04 ERROR 404: Not Found.

按照如下方法，将其全部进行修改。

训练

torchpack dist-run -np 1 python tools/train.py configs/nuscenes/det/transfusion/secfpn/camera+lidar/swint_v0p075/convfuser.yaml --model.encoders.camera.backbone.init_cfg.checkpoint pretrained/swint-nuimages-pretrained.pth --load_from pretrained/lidar-only-det.pth

测试

# pretrained
torchpack dist-run -np 1 python tools/test.py configs/nuscenes/det/transfusion/secfpn/camera+lidar/swint_v0p075/convfuser.yaml pretrained/bevfusion-det.pth --eval bbox

# customer
torchpack dist-run -np 1 python tools/test.py train_result/configs.yaml train_result/latest.pth --eval bbox --out box.pkl

可视化

# pretrained
torchpack dist-run -np 1 python tools/visualize.py configs/nuscenes/det/transfusion/secfpn/camera+lidar/swint_v0p075/convfuser.yaml --mode pred --checkpoint pretrained/bevfusion-det.pth --bbox-score 0.2 --out-dir vis_result

# customer
torchpack dist-run -np 1 python tools/visualize.py
train_result/configs.yaml --mode gt --checkpoint train_result/latest.pth --bbox-score 0.2 --out-dir vis_result