论文笔记_S2D.52_CMRNet++_运行记录

一时失言乱红尘 2022-12-27 14:13 204阅读 0赞

**目录**

1 环境搭建

2 需要的文件

2.1 pykitti中的默认的文件位置

2.2 preprocess | 运行 kitti\_maps.py 需要的文件

2.3 preprocess | 运行 kitti\_maps.py 生成的文件

2.4 Training | 运行 main\_visibility\_CALIB.py 需要的文件

2.5 Evaluation | 运行 evaluate\_iterative\_single\_CALIB.py 需要的文件

3 运行命令

--------------------

*  [论文笔记\_S2D.52\_CMRNet++：在激光雷达地图中进行内参未知的相机的单目视觉定位][S2D.52_CMRNet]
 *  开源代码位置: [https://github.com/cattaneod/CMRNet][https_github.com_cattaneod_CMRNet]

# 1 环境搭建 #

*  1. conda create -n py36env\_name python=3.6
 *  2. https://github.com/cattaneod/CMRNet 在项目文件夹下执行pip install -r requirements.txt

# 2 需要的文件 #

(设置 "data\_folder=./KITTI\_ODOMETRY/sequences"时)-----

## 2.1 pykitti中的默认的文件位置 ##

*  self.sequence\_path: ./KITTI\_ODOMETRY/sequences/00/
 *  对应pose\_file: ./KITTI\_ODOMETRY/poses/00 .txt
 *  self.velo\_files: ./KITTI\_ODOMETRY/sequences/00/velodyne/\*.bin

## 2.2 preprocess | 运行 kitti\_maps.py 需要的文件 ##

*  矫正文件: ./KITTI\_ODOMETRY/sequences/00/calib.txt
 *  激光数据: /KITTI\_ODOMETRY/sequences/00/velodyne/\*.bin
 *  位姿数据: ./data/kitti-00.csv

## 2.3 preprocess | 运行 kitti\_maps.py 生成的文件 ##

*  单个局部地图: ./KITTI\_ODOMETRY/sequences/00/local\_maps\_0.1/000xxx.h5
 *  序列的全局地图: ./map-00\_0.1\_0-4541.pcd

## 2.4 Training | 运行 main\_visibility\_CALIB.py 需要的文件 ##

*  位姿数据: ./KITTI\_ODOMETRY/sequences/00/poses.csv
 *  图像文件: ./KITTI\_ODOMETRY/sequences/00/image\_2/00xxx.png
 *  对应的h5文件: ./KITTI\_ODOMETRY/sequences/00/velodyne/00xxx.h5
 *  运行结果: 生成文件./checkpoints/kitti/00/checkpoint\_r10.00\_t2.00\_e2\_1.200.tar

## 2.5 Evaluation | 运行 evaluate\_iterative\_single\_CALIB.py 需要的文件 ##

*  ./KITTI\_ODOMETRY/sequences/test\_RT\_seq00\_10.00\_2.00.csv
 *  ./checkpoints/kitti/00/checkpoint\_r10.00\_t2.00\_e2\_1.200.tar

# 3 运行命令 #

预处理 preprocess:

python preprocess/kitti_maps.py --sequence 00 --kitti_folder ./KITTI_ODOMETRY/sequences

Training:

python main_visibility_CALIB.py with batch_size=24 data_folder=./KITTI_ODOMETRY/sequences epochs=300 max_r=10 max_t=2 BASE_LEARNING_RATE=0.0001 savemodel=./checkpoints/ test_sequence=00

Evaluation:

python evaluate_iterative_single_CALIB.py with test_sequence=00 maps_folder=local_maps data_folder=./KITTI_ODOMETRY/sequences weight="['./checkpoints/kitti/00/checkpoint_r10.00_t2.00_e2_1.200.tar']"

[S2D.52_CMRNet]: https://blog.csdn.net/shyjhyp11/article/details/110391044
[https_github.com_cattaneod_CMRNet]: https://github.com/cattaneod/CMRNet