# bevfusion

**Repository Path**: ericzhou2018/bevfusion

## Basic Information

- **Project Name**: bevfusion
- **Description**: bevfusion稳定版本：db75150
- **Primary Language**: Unknown
- **License**: Apache-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2026-02-25
- **Last Updated**: 2026-02-25

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# bevfusion

1. 下载项目

`git clone https://gitee.com/sjj668/bevfusion.git`

2. 创建环境

`conda create -n bevfusion python=3.8` 

3. 环境搭建
cuda：11.3
```
sudo apt-get update && apt-get install wget -yq
sudo apt-get install build-essential g++ gcc -y
sudo apt-get install libgl1-mesa-glx libglib2.0-0 -y
sudo apt-get install openmpi-bin openmpi-common libopenmpi-dev libgtk2.0-dev git -y


conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=11.3 -c pytorch
pip config set global.index-url https://mirrors.bfsu.edu.cn/pypi/web/simple
pip install Pillow==8.4.0
pip install tqdm
pip install torchpack
pip install mmcv==1.4.0
pip install mmcv-full==1.4.0 -f https://download.openmmlab.com/mmcv/dist/cu113/torch1.10/index.html
pip install mmdet==2.20.0
pip install nuscenes-devkit
pip install mpi4py==3.0.3
pip install numba==0.48.0
pip install numpy==1.19.5
pip install setuptools==59.5.0
pip install yapf==0.40.1
pip install spconv-cu113
```
4. 代码编译（注意新项目别忘记进行编译）

`python setup.py develop`

5. 数据集准备

```
#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
```
数据处理

```
#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
```
7. 数据训练

```
# train
## camera
torchpack dist-run -np 2 python tools/train.py configs/nuscenes/det/centerhead/lssfpn/camera/256x704/swint/default.yaml --model.encoders.camera.backbone.init_cfg.checkpoint pretrained/swint-nuimages-pretrained.pth --run-dir output/camera_result

## camera+lidar
torchpack dist-run -np 4 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 --run-dir output/camera+lidar_result

## lidar
torchpack dist-run -np 4 python tools/train.py configs/nuscenes/det/transfusion/secfpn/lidar/voxelnet_0p075.yaml --run-dir output/lidar_result_0106

# test

torchpack dist-run -np 1 python tools/test.py configs/nuscenes/det/transfusion/secfpn/camera+lidar/swint_v0p075/convfuser.yaml output/camera+lidar_result/latest.pth --eval bbox --show-dir output/test/camera+lidar_result

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

### test lidar
torchpack dist-run -np 4 python tools/test.py configs/nuscenes/det/transfusion/secfpn/lidar/voxelnet_0p075.yaml  pretrained/lidar-only-det.pth --eval map

### test camera
torchpack dist-run -np 4 python tools/test.py configs/nuscenes/det/centerhead/lssfpn/camera/256x704/swint/default.yaml pretrained/camera-only-det.pth --eval map


# visualize
torchpack dist-run -np 4 python tools/visualize.py /home/nvidia/projects/bevfusion/configs.yaml --mode gt --checkpoint /home/nvidia/projects/bevfusion/latest.pth --bbox-score 0.5 --out-dir vis_result

torchpack dist-run -np 4 python tools/visualize.py /home/nvidia/Projects/bevfusion/output/train/camera+lidar_result_1023/configs.yaml --mode pred --checkpoint /home/nvidia/Projects/bevfusion/output/train/camera+lidar_result_1023/latest.pth --bbox-score 0.5 --out-dir vis_result_pred


nohup 训练指令 >> ./work_dirs/train_output.log 2>&1 &

```


注意减小batchsize后学习率也要相应减小 反之亦然