一、参考资料

KITTI数据集

二、重要说明

在monodepth2复现过程中，训练数据的格式是：

# kitti数据集的真实路径
/media/yoyo/新加卷/Downloads/kitti_data# 训练数据集的路径
/home/yoyo/MyDocuments/PyProjects/monodepth2/kitti_data/2011_09_28/2011_09_28_drive_0016_sync# 训练数据集创建软链接
sudo ln -s /media/yoyo/新加卷/Downloads/kitti_data /home/yoyo/MyDocuments/PyProjects/monodepth2/kitti_data

monodepth2训练需要的 KITTIRaw 数据集

2011_09_26_calib.zip
2011_09_26_drive_0001/2011_09_26_drive_0001_sync.zip
2011_09_26_drive_0002/2011_09_26_drive_0002_sync.zip
2011_09_26_drive_0005/2011_09_26_drive_0005_sync.zip
2011_09_26_drive_0009/2011_09_26_drive_0009_sync.zip
2011_09_26_drive_0011/2011_09_26_drive_0011_sync.zip
2011_09_26_drive_0013/2011_09_26_drive_0013_sync.zip
2011_09_26_drive_0014/2011_09_26_drive_0014_sync.zip
2011_09_26_drive_0015/2011_09_26_drive_0015_sync.zip
2011_09_26_drive_0017/2011_09_26_drive_0017_sync.zip
2011_09_26_drive_0018/2011_09_26_drive_0018_sync.zip
2011_09_26_drive_0019/2011_09_26_drive_0019_sync.zip
2011_09_26_drive_0020/2011_09_26_drive_0020_sync.zip
2011_09_26_drive_0022/2011_09_26_drive_0022_sync.zip
2011_09_26_drive_0023/2011_09_26_drive_0023_sync.zip
2011_09_26_drive_0027/2011_09_26_drive_0027_sync.zip
2011_09_26_drive_0028/2011_09_26_drive_0028_sync.zip
2011_09_26_drive_0029/2011_09_26_drive_0029_sync.zip
2011_09_26_drive_0032/2011_09_26_drive_0032_sync.zip
2011_09_26_drive_0035/2011_09_26_drive_0035_sync.zip
2011_09_26_drive_0036/2011_09_26_drive_0036_sync.zip
2011_09_26_drive_0039/2011_09_26_drive_0039_sync.zip
2011_09_26_drive_0046/2011_09_26_drive_0046_sync.zip
2011_09_26_drive_0048/2011_09_26_drive_0048_sync.zip
2011_09_26_drive_0051/2011_09_26_drive_0051_sync.zip
2011_09_26_drive_0052/2011_09_26_drive_0052_sync.zip
2011_09_26_drive_0056/2011_09_26_drive_0056_sync.zip
2011_09_26_drive_0057/2011_09_26_drive_0057_sync.zip
2011_09_26_drive_0059/2011_09_26_drive_0059_sync.zip
2011_09_26_drive_0060/2011_09_26_drive_0060_sync.zip
2011_09_26_drive_0061/2011_09_26_drive_0061_sync.zip
2011_09_26_drive_0064/2011_09_26_drive_0064_sync.zip
2011_09_26_drive_0070/2011_09_26_drive_0070_sync.zip
2011_09_26_drive_0079/2011_09_26_drive_0079_sync.zip
2011_09_26_drive_0084/2011_09_26_drive_0084_sync.zip
2011_09_26_drive_0086/2011_09_26_drive_0086_sync.zip
2011_09_26_drive_0087/2011_09_26_drive_0087_sync.zip
2011_09_26_drive_0091/2011_09_26_drive_0091_sync.zip
2011_09_26_drive_0093/2011_09_26_drive_0093_sync.zip
2011_09_26_drive_0095/2011_09_26_drive_0095_sync.zip
2011_09_26_drive_0096/2011_09_26_drive_0096_sync.zip
2011_09_26_drive_0101/2011_09_26_drive_0101_sync.zip
2011_09_26_drive_0104/2011_09_26_drive_0104_sync.zip
2011_09_26_drive_0106/2011_09_26_drive_0106_sync.zip
2011_09_26_drive_0113/2011_09_26_drive_0113_sync.zip
2011_09_26_drive_0117/2011_09_26_drive_0117_sync.zip
2011_09_28_calib.zip
2011_09_28_drive_0001/2011_09_28_drive_0001_sync.zip
2011_09_28_drive_0002/2011_09_28_drive_0002_sync.zip
2011_09_29_calib.zip
2011_09_29_drive_0004/2011_09_29_drive_0004_sync.zip
2011_09_29_drive_0026/2011_09_29_drive_0026_sync.zip
2011_09_29_drive_0071/2011_09_29_drive_0071_sync.zip
2011_09_30_calib.zip
2011_09_30_drive_0016/2011_09_30_drive_0016_sync.zip
2011_09_30_drive_0018/2011_09_30_drive_0018_sync.zip
2011_09_30_drive_0020/2011_09_30_drive_0020_sync.zip
2011_09_30_drive_0027/2011_09_30_drive_0027_sync.zip
2011_09_30_drive_0028/2011_09_30_drive_0028_sync.zip
2011_09_30_drive_0033/2011_09_30_drive_0033_sync.zip
2011_09_30_drive_0034/2011_09_30_drive_0034_sync.zip
2011_10_03_calib.zip
2011_10_03_drive_0027/2011_10_03_drive_0027_sync.zip
2011_10_03_drive_0034/2011_10_03_drive_0034_sync.zip
2011_10_03_drive_0042/2011_10_03_drive_0042_sync.zip
2011_10_03_drive_0047/2011_10_03_drive_0047_sync.zip

三、关键步骤

下载格物钛数据集 - KITTIRaw
其他下载方式，参考博客 KITTI数据集

.
├── KITTIRaw_calibration.zip
├── KITTIRaw_campus.zip
├── KITTIRaw_city.zip
├── KITTIRaw_person.zip
├── KITTIRaw_residential.zip
└── KITTIRaw_road.zip
...

解压zip文件

每一个zip文件都很大，需要逐个解压

# 以KITTIRaw_campus.zip为例，进行解压
unzip KITTIRaw_campus.zip

.
├── campus
│   ├── 2011_09_28_drive_0016
│   ├── 2011_09_28_drive_0021
│   ├── 2011_09_28_drive_0034
│   ├── 2011_09_28_drive_0035
│   ├── 2011_09_28_drive_0037
│   ├── 2011_09_28_drive_0038
│   ├── 2011_09_28_drive_0039
│   ├── 2011_09_28_drive_0043
│   ├── 2011_09_28_drive_0045
│   └── 2011_09_28_drive_0047
...

# 以 2011_09_28_drive_0016 为例
# /media/yoyo/新加卷/Downloads/campus/2011_09_28_drive_00162011_09_28_drive_0016
├── 2011_09_28_calib.zip
├── 2011_09_28_drive_0016_extract.zip
└── 2011_09_28_drive_0016_sync.zip
...

#  2011_09_28_drive_0016_sync.zip 文件
unzip 2011_09_28_drive_0016_sync.zip

# /media/yoyo/新加卷/Downloads/campus/2011_09_28_drive_0016/2011_09_28/2011_09_28_drive_0016_sync
.
├── 2011_09_28_drive_0016
│   ├── 2011_09_28
│   │   └── 2011_09_28_drive_0016_sync
│   │       ├── image_00
│   │       ├── image_01
│   │       ├── image_02
│   │       ├── image_03
│   │       ├── oxts
│   │       └── velodyne_points
...

从KITTIRaw原始zip压缩文件中，挑选出需要的zip文件
用python解压所有的zip文件

import os
import zipfile
import shutildir_one_zip_example = '/home/yoyo/MyDocuments/PyProjects/202109/20210905/2011_09_26_drive_0001_sync.zip'# /media/yoyo/新加卷/Downloads/kitti_data
'''
├── kitti_data
│   ├── 2011_09_26_calib.zip
│   ├── 2011_09_26_drive_0001_sync.zip
│   ├── 2011_09_26_drive_0002_sync.zip
│   ├── 2011_09_26_drive_0005_sync.zip
│   ├── 2011_09_26_drive_0009_sync.zip
│   ├── 2011_09_26_drive_0011_sync.zip
│   ├── 2011_09_26_drive_0013_sync.zip
│   ├── 2011_09_26_drive_0014_sync.zip
│   ├── 2011_09_26_drive_0015_sync.zip
│   ├── 2011_09_26_drive_0017_sync.zip
│   ├── 2011_09_26_drive_0018_sync.zip
│   ├── 2011_09_26_drive_0019_sync.zip
│   ├── 2011_09_26_drive_0020_sync.zip
│   ├── 2011_09_26_drive_0022_sync.zip
│   ├── 2011_09_26_drive_0023_sync.zip
│   ├── 2011_09_26_drive_0027_sync.zip
│   ├── 2011_09_26_drive_0028_sync.zip
│   ├── 2011_09_26_drive_0029_sync.zip
│   ├── 2011_09_26_drive_0032_sync.zip
│   ├── 2011_09_26_drive_0035_sync.zip
│   ├── 2011_09_26_drive_0036_sync.zip
│   ├── 2011_09_26_drive_0039_sync.zip
│   ├── 2011_09_26_drive_0046_sync.zip
│   ├── 2011_09_26_drive_0048_sync.zip
│   ├── 2011_09_26_drive_0051_sync.zip
│   ├── 2011_09_26_drive_0052_sync.zip
│   ├── 2011_09_26_drive_0056_sync.zip
│   ├── 2011_09_26_drive_0057_sync.zip
│   ├── 2011_09_26_drive_0059_sync.zip
│   ├── 2011_09_26_drive_0060_sync.zip
│   ├── 2011_09_26_drive_0061_sync.zip
│   ├── 2011_09_26_drive_0064_sync.zip
│   ├── 2011_09_26_drive_0070_sync.zip
│   ├── 2011_09_26_drive_0079_sync.zip
│   ├── 2011_09_26_drive_0084_sync.zip
│   ├── 2011_09_26_drive_0086_sync.zip
│   ├── 2011_09_26_drive_0087_sync.zip
│   ├── 2011_09_26_drive_0091_sync.zip
│   ├── 2011_09_26_drive_0093_sync.zip
│   ├── 2011_09_26_drive_0095_sync.zip
│   ├── 2011_09_26_drive_0096_sync.zip
│   ├── 2011_09_26_drive_0101_sync.zip
│   ├── 2011_09_26_drive_0104_sync.zip
│   ├── 2011_09_26_drive_0106_sync.zip
│   ├── 2011_09_26_drive_0113_sync.zip
│   ├── 2011_09_26_drive_0117_sync.zip
│   ├── 2011_09_28_calib.zip
│   ├── 2011_09_28_drive_0001_sync.zip
│   ├── 2011_09_28_drive_0002_sync.zip
│   ├── 2011_09_29_calib.zip
│   ├── 2011_09_29_drive_0004_sync.zip
│   ├── 2011_09_29_drive_0026_sync.zip
│   ├── 2011_09_29_drive_0071_sync.zip
│   ├── 2011_09_30_calib.zip
│   ├── 2011_09_30_drive_0016_sync.zip
│   ├── 2011_09_30_drive_0018_sync.zip
│   ├── 2011_09_30_drive_0020_sync.zip
│   ├── 2011_09_30_drive_0027_sync.zip
│   ├── 2011_09_30_drive_0028_sync.zip
│   ├── 2011_09_30_drive_0033_sync.zip
│   ├── 2011_09_30_drive_0034_sync.zip
│   ├── 2011_10_03_calib.zip
│   ├── 2011_10_03_drive_0027_sync.zip
│   ├── 2011_10_03_drive_0034_sync.zip
│   ├── 2011_10_03_drive_0042_sync.zip
│   ├── 2011_10_03_drive_0047_sync.zip
'''def unzip_one_file(dir_one_zip=dir_one_zip_example):'''解压单个zip文件/media/yoyo/新加卷/Downloads/campus/2011_09_28_drive_0016/2011_09_28_drive_0016_sync.zip:param dir_one_zip: zip的路径:return:'''dir_path = os.path.split(dir_one_zip)[0]if not zipfile.is_zipfile(dir_one_zip):  # 如果不可以解压print(dir_one_zip, 'is not a zipfile!')  # 打印不可解压的zip文件returnwith zipfile.ZipFile(dir_one_zip) as zf:  # 打开zip压缩文件if 'drive' in dir_one_zip:  # 如果类似于 2011_09_28_drive_0016_sync.zip 的文件dir_date_first = dir_one_zip.split('_drive')[0]  # /media/yoyo/新加卷/Downloads/campus/2011_09_28_drive_0016/2011_09_28prefix_name = os.path.split(dir_date_first)[-1]  # 2011_09_28if not os.path.exists(dir_date_first):  # 如果文件夹不存在，创建文件夹os.mkdir(dir_date_first)temp_dir = os.path.join(dir_path, 'temp_dir')  # 存放解压文件的临时文件夹dir_date_second = os.path.join(temp_dir, prefix_name)if not os.path.exists(temp_dir):  # 创建临时文件夹os.mkdir(temp_dir)zf.extractall(temp_dir)  # 解压到临时文件夹obj_dir = os.listdir(dir_date_second)if len(obj_dir) == 1:src_dir = os.path.join(dir_date_second, obj_dir[0])shutil.move(src_dir, dir_date_first)  # 移动解压后的文件夹shutil.rmtree(dir_date_second)  # 删除解压后的文件夹def unzip_all_file(dir_all_zip):'''解压zip根目录的所有zip文件/media/yoyo/新加卷/Downloads/kitti_data:param dir_all_zip: zip根目录:return:'''dir_zip_list = os.listdir(dir_all_zip)for idx, zip_item in enumerate(dir_zip_list):dir_one_zip = os.path.join(dir_all_zip, zip_item)unzip_one_file(dir_one_zip)  # 解压单个zip文件print(idx+1, dir_one_zip)if __name__ == '__main__':dir_all_zip = '/media/yoyo/新加卷/Downloads/kitti_data'  # zip文件根目录，待解压unzip_all_file(dir_all_zip)

复现monodepth2之KITTI数据集准备相关推荐

【图像语义分割】DeepLabv3+(Pytorch版) 源码复现—Kitti数据集
自己研究课题涉及到图像语义分割,所以最近复现了Deeplabv3+,在KITTI数据集上进行实验,kitti的语义分割数据集是按照cityspace的数据集制作的所有可以用cityspaces的训练模 ...
基于KITTI数据集的无人驾驶感知与传感器融合实现—（0）—照猫画虎
基于KITTI数据集的无人驾驶感知与传感器融合实现-(0)-照猫画虎学习前言链接项目介绍项目效果逻辑框架图详细步骤学习前言最近自动驾驶蛮火的,我自己看这玩意也想弄弄,然后在百度AI ...
深度估计自监督模型monodepth2在自己数据集的实战——单卡/多卡训练、推理、Onnx转换和量化指标评估
本文详细介绍monodepth2模型在自己数据集的实战方法,包括单卡/多卡训练.推理.Onnx转换和量化评估等,关于理论部分请参见另一篇博客:深度估计自监督模型monodepth2论文总结和源码分析 ...
双目测距系列（七）monodepth2训练前数据集准备过程的简析
前言上一篇文章说过,monodepth2模型有三种训练方式.针对我们的双目场景,准备使用stereo training方法. monodepth2的训练入口函数在train.py中,如下图所示. 总 ...
KITTI数据集上MaskRCNN检测效果示例
KITTI数据集上MaskRCNN检测效果示例在Semantic Instance Segmentation Evaluation中,MaskRCNN性能效果排名第一. Test Image 0 I ...
jetson nano 用 tensorrt 运行 nanodet(kitti数据集)
题目是目标,我们先一步一步来做,第一步是训练神经网络, 我用的是 kitti 数据集,训练顺序为,第一步,拿到kitti数据集,第二步,把kitti数据集修改为voc数据集格式,第三步,修改配置文件进 ...
「Self-driving: KITTI」KITTI数据集简介
背景 KITTI数据集建立前,前人的实验在米德尔伯里等既定基准上排名靠前的方法在从实验室转移到现实世界时表现低于平均水平. 除了向社区提供具现实世界数据以减少实验偏差,KITTI在补充现有基准的情况下 ...
kitti数据集_KITTI数据集激光雷达坐标系下的里程计真值
由KITTI数据集Odometry模块的devkit_odometrydevkitreadme.txt所述,KITII数据集提供里程计的真值是在左相机表坐标系下的,并没有提供激光雷达坐标系下的真值.因 ...
多相机坐标转换_使用KITTI数据集实现坐标转换
本文的主要内容,是讨论如何将KITTI激光雷达的点云数据,映射到图像中去. KITTI数据集是怎么采集的下面这幅图是车的俯视图,可以看到,用到了2个彩色相机,2个灰度相机,1个激光雷达.1个GPS/ ...
KITTI数据集的使用——雷达与相机的数据融合
目录目的如何实现 kitti数据集简介 kitti数据集的raw_data 利用kitti提供的devkit以及相应数据集的calib文件解读calib文件夹解读devkit 目的使用雷达点 ...

复现monodepth2之KITTI数据集准备

一、参考资料

二、重要说明

三、关键步骤

复现monodepth2之KITTI数据集准备相关推荐

最新文章

热门文章