当前位置： 首页 > news >正文

MatterPort3D 数据集 | 简介 | 多途径下载

news 2025/7/23 5:03:59

Matterport3D，简称MP3D，是室内场景的一个大规模 RGB-D 数据集。可用于机器人具身导航算法开发。

包含 90 个建筑规模场景的 194,400 张RGB-D图像，以及10,800个全景视图。

数据集提供了表面重建、相机姿态以及二维和三维语义分割的注释。

支持各种监督和自监督计算机视觉任务，包括关键点匹配、视图重叠预测、基于颜色的法线预测、语义分割和场景分类。

项目地址：https://niessner.github.io/Matterport/

代码地址：https://github.com/niessner/Matterport

论文地址：Matterport3D: Learning from RGB-D Data in Indoor Environments

1、数据集下载（官方）

该数据集包含多种类型的标注：彩色和深度图像、相机姿态、带纹理的 3D 网格、建筑平面图和区域标注、对象实例语义标注。

数据集的格式参考：https://github.com/niessner/Matterport/blob/master/data_organization.md

需要填写并签署使用条款协议表格，然后将其发送至matterport3d@googlegroups.com以请求访问数据集。

2、Habitat-Lab 提供示例下载

Scenes datasets 如下表格所示：

Scenes modelsExtract pathArchive size
Habitat test scenesdata/scene_datasets/habitat-test-scenes/{scene}.glb89 MB
ReplicaCADdata/scene_datasets/replica_cad/configs/scenes/{scene}.scene_instance.json123 MB
HM3Ddata/scene_datasets/hm3d/{split}/00\d\d\d-{scene}/{scene}.basis.glb130 GB
Gibsondata/scene_datasets/gibson/{scene}.glb1.5 GB
MatterPort3Ddata/scene_datasets/mp3d/{scene}/{scene}.glb15 GB
 HSSD-Habitatdata/scene_datasets/hssd-hab/scenes/{scene}.scene_instance.jsonXXXXX GB
AI2-THOR-Habitatdata/scene_datasets/ai2thor-hab/ai2thor-hab/configs/scenes/{DATASET}/{scene}.scene_instance.jsonXXXXX GB

点击“Matterport3D”后，能看到下载命令：

python -m habitat_sim.utils.datasets_download --uids mp3d_example_scene --data-path data/

但是这只是一个示例场景，用于在 Habitat-sim 中执行单元测试，无法下载完整的“Matterport3D”数据。

Task datasets 如下表格所示：

TaskScenesLinkExtract pathConfig to useArchive size
Rearrange PickReplicaCADrearrange_pick_replica_cad_v0.zipdata/datasets/rearrange_pick/replica_cad/v0/datasets/rearrangepick/replica_cad.yaml11 MB
Point goal navigationGibsonpointnav_gibson_v1.zipdata/datasets/pointnav/gibson/v1/datasets/pointnav/gibson.yaml385 MB
Point goal navigationGibson 0+ (train)pointnav_gibson_0_plus_v1.zipdata/datasets/pointnav/gibson/v1/datasets/pointnav/gibson_0_plus.yaml321 MB
Point goal navigation corresponding to Sim2LoCoBot experiment configurationGibsonpointnav_gibson_v2.zipdata/datasets/pointnav/gibson/v2/datasets/pointnav/gibson_v2.yaml274 MB
Point goal navigationMatterPort3Dpointnav_mp3d_v1.zipdata/datasets/pointnav/mp3d/v1/datasets/pointnav/mp3d.yaml400 MB
Point goal navigationHM3Dpointnav_hm3d_v1.zipdata/datasets/pointnav/hm3d/v1/datasets/pointnav/hm3d.yaml992 MB
Object goal navigationMatterPort3Dobjectnav_mp3d_v1.zipdata/datasets/objectnav/mp3d/v1/datasets/objectnav/mp3d.yaml170 MB
Object goal navigationHM3DSem-v0.1objectnav_hm3d_v1.zipdata/datasets/objectnav/hm3d/v1/datasets/objectnav/hm3d.yaml154 MB
Object goal navigationHM3DSem-v0.2objectnav_hm3d_v2.zipdata/datasets/objectnav/hm3d/v2/datasets/objectnav/hm3d_v2.yaml245 MB
Object goal navigationHSSDobjectnav_hssd_v0.2.3.zipdata/datasets/objectnav/hssd-habdatasets/objectnav/hssd-hab.yaml206 MB
Object goal navigationProcTHOR-habobjectnav_procthor-hab.zipdata/datasets/objectnav/procthor-habdatasets/objectnav/procthor-hab.yaml755 MB
Embodied Question AnsweringMatterPort3Deqa_mp3d_v1.zipdata/datasets/eqa/mp3d/v1/datasets/eqa/mp3d.yaml44 MB
Visual Language NavigationMatterPort3Dvln_r2r_mp3d_v1.zipdata/datasets/vln/mp3d/r2r/v1datasets/vln/mp3d_r2r.yaml2.7 MB
Instance image goal navigationHM3DSem-v0.1instance_imagenav_hm3d_v1.zipdata/datasets/instance_imagenav/hm3d/v1/datasets/instance_imagenav/hm3d_v1.yaml303 MB
Instance image goal navigationHM3DSem-v0.2instance_imagenav_hm3d_v2.zipdata/datasets/instance_imagenav/hm3d/v2/datasets/instance_imagenav/hm3d_v2.yaml518 MB
Instance image goal navigationHM3DSem-v0.2instance_imagenav_hm3d_v3.zipdata/datasets/instance_imagenav/hm3d/v3/datasets/instance_imagenav/hm3d_v3.yaml517 MB
Image goal navigationGibsonpointnav_gibson_v1.zipdata/datasets/pointnav/gibson/v1/datasets/imagenav/gibson.yaml385 MB
Image goal navigationMatterPort3Dpointnav_mp3d_v1.zipdata/datasets/pointnav/mp3d/v1/datasets/imagenav/mp3d.yaml400 MB

3、批量下载

首先创建一个Conda环境，名字为mp3d，python版本为3.9

进入mp3d环境

conda create -n mp3d python=3.9
conda activate mp3d

然后安装requests依赖库

pip install requests

编写批量下载的代码：

#!/usr/bin/env python
# -*- coding: utf-8 -*-import argparse
import collections
import os
import tempfile
import urllib
from urllib import requestimport requests
from time import sleep
from requests.exceptions import ConnectionError, ChunkedEncodingError, RequestException
import sysBASE_URL = 'http://kaldir.vc.in.tum.de/matterport/'
RELEASE = 'v1/scans'
RELEASE_TASKS = 'v1/tasks/'
RELEASE_SIZE = '1.3TB'
TOS_URL = BASE_URL + 'MP_TOS.pdf'
FILETYPES = ['cameras','matterport_camera_intrinsics','matterport_camera_poses','matterport_color_images','matterport_depth_images','matterport_hdr_images','matterport_mesh','matterport_skybox_images','undistorted_camera_parameters','undistorted_color_images','undistorted_depth_images','undistorted_normal_images','house_segmentations','region_segmentations','image_overlap_data','poisson_meshes','sens'
]
TASK_FILES = {'keypoint_matching_data': ['keypoint_matching/data.zip'],'keypoint_matching_models': ['keypoint_matching/models.zip'],'surface_normal_data': ['surface_normal/data_list.zip'],'surface_normal_models': ['surface_normal/models.zip'],'region_classification_data': ['region_classification/data.zip'],'region_classification_models': ['region_classification/models.zip'],'semantic_voxel_label_data': ['semantic_voxel_label/data.zip'],'semantic_voxel_label_models': ['semantic_voxel_label/models.zip'],'minos': ['mp3d_minos.zip'],'gibson': ['mp3d_for_gibson.tar.gz'],'habitat': ['mp3d_habitat.zip'],'pixelsynth': ['mp3d_pixelsynth.zip'],'igibson': ['mp3d_for_igibson.zip'],'mp360': ['mp3d_360/data_00.zip', 'mp3d_360/data_01.zip', 'mp3d_360/data_02.zip', 'mp3d_360/data_03.zip', 'mp3d_360/data_04.zip', 'mp3d_360/data_05.zip', 'mp3d_360/data_06.zip']
}def get_release_scans(release_file):# scan_lines = urllib.urlopen(release_file)scan_lines = request.urlopen(release_file)scans = []for scan_line in scan_lines:scan_line = str(scan_line, 'utf-8')scan_id = scan_line.rstrip('\n')scans.append(scan_id)return scansdef download_release(release_scans, out_dir, file_types):print('Downloading MP release to ' + out_dir + '...')for scan_id in release_scans:scan_out_dir = os.path.join(out_dir, scan_id)download_scan(scan_id, scan_out_dir, file_types)print('Downloaded MP release.')def download_file(url, out_file, max_retries=5, chunk_size=1024*1024):"""下载单个文件，支持断点续传和自动重试，并打印实时下载进度。兼容 Python2/3。"""out_dir = os.path.dirname(out_file)if not os.path.isdir(out_dir):try:os.makedirs(out_dir)except OSError:pass# 获取文件总大小head = requests.head(url, allow_redirects=True)if head.status_code != 200:raise IOError("无法获取文件大小: {0}，状态码 {1}".format(url, head.status_code))total_size = int(head.headers.get('Content-Length', 0))# 计算已下载起点resume = 0if os.path.exists(out_file):resume = os.path.getsize(out_file)if resume >= total_size:print("跳过已存在文件 %s" % out_file)returnprint("开始下载 %s (%0.2f MB)" % (out_file, total_size / 1024.0**2))retries = 0last_print = 0while resume < total_size and retries <= max_retries:headers = {"Range": "bytes=%d-%d" % (resume, total_size - 1)}try:r = requests.get(url, headers=headers, stream=True, timeout=30)r.raise_for_status()fh, tmp_path = tempfile.mkstemp(dir=out_dir)with os.fdopen(fh, "wb") as tmpf:for chunk in r.iter_content(chunk_size=chunk_size):if not chunk:continuetmpf.write(chunk)resume += len(chunk)# 每下载 5% 或最后一块时打印一次进度percent = int(resume * 100 / total_size)if percent - last_print >= 5 or resume == total_size:sys.stdout.write("\r下载进度: %3d%% (%0.2f/%0.2f MB)" % (percent,resume  / 1024.0**2,total_size / 1024.0**2))sys.stdout.flush()last_print = percent# 将本次下载数据追加到目标文件with open(tmp_path, "rb") as tmpf, open(out_file, "ab") as outf:outf.write(tmpf.read())os.remove(tmp_path)r.close()break# except (requests.ConnectionError, requests.ChunkedEncodingError, IOError) as e:except (ConnectionError, ChunkedEncodingError, IOError) as e:retries += 1wait = 2 ** retriesprint("\n[重试 %d/%d] 已下载 %0.2f MB，等待 %d 秒后重试..." % (retries, max_retries,resume / 1024.0**2,wait))sleep(wait)if resume < total_size:raise IOError("下载失败：只获取到 %d / %d 字节" % (resume, total_size))# 下载完成后换行sys.stdout.write("\n下载完成：%s\n" % out_file)def download_scan(scan_id, out_dir, file_types):print('Downloading MP scan ' + scan_id + ' ...')if not os.path.isdir(out_dir):os.makedirs(out_dir)for ft in file_types:url = BASE_URL + RELEASE + '/' + scan_id + '/' + ft + '.zip'out_file = out_dir + '/' + ft + '.zip'download_file(url, out_file)print('Downloaded scan ' + scan_id)def download_task_data(task_data, out_dir):print('Downloading MP task data for ' + str(task_data) + ' ...')for task_data_id in task_data:if task_data_id in TASK_FILES:file = TASK_FILES[task_data_id]for filepart in file:url = BASE_URL + RELEASE_TASKS + '/' + filepartlocalpath = os.path.join(out_dir, filepart)localdir = os.path.dirname(localpath)if not os.path.isdir(localdir):os.makedirs(localdir)download_file(url, localpath)print('Downloaded task data ' + task_data_id)def main():parser = argparse.ArgumentParser(description='''Downloads MP public data release.Example invocation:python download_mp.py -o base_dir --id ALL --type object_segmentations --task_data semantic_voxel_label_data semantic_voxel_label_modelsThe -o argument is required and specifies the base_dir local directory.After download base_dir/v1/scans is populated with scan data, and base_dir/v1/tasks is populated with task data.Unzip scan files from base_dir/v1/scans and task files from base_dir/v1/tasks/task_name.The --type argument is optional (all data types are downloaded if unspecified).The --id ALL argument will download all house data. Use --id house_id to download specific house data.The --task_data argument is optional and will download task data and model files.''',formatter_class=argparse.RawTextHelpFormatter)parser.add_argument('-o', '--out_dir', required=True, help='directory in which to download')parser.add_argument('--task_data', default=[], nargs='+', help='task data files to download. Any of: ' + ','.join(TASK_FILES.keys()))parser.add_argument('--id', default='ALL', help='specific scan id to download or ALL to download entire dataset')parser.add_argument('--type', nargs='+', help='specific file types to download. Any of: ' + ','.join(FILETYPES))args = parser.parse_args()release_file = BASE_URL + RELEASE + '.txt'release_scans = get_release_scans(release_file)file_types = FILETYPES# download task dataif args.task_data:if set(args.task_data) & set(TASK_FILES.keys()):  # download task dataout_dir = os.path.join(args.out_dir, RELEASE_TASKS)download_task_data(args.task_data, out_dir)else:print('ERROR: Unrecognized task data id: ' + args.task_data)print('Done downloading task_data for ' + str(args.task_data))# key = raw_input('Press any key to continue on to main dataset download, or CTRL-C to exit.')key = input('Press any key to continue on to main dataset download, or CTRL-C to exit.')# download specific file types?if args.type:if not set(args.type) & set(FILETYPES):# print('ERROR: Invalid file type: ' + file_type)print('ERROR: Invalid file type: ' + file_types)returnfile_types = args.typeif args.id and args.id != 'ALL':  # download single scanscan_id = args.idif scan_id not in release_scans:print('ERROR: Invalid scan id: ' + scan_id)else:out_dir = os.path.join(args.out_dir, RELEASE, scan_id)download_scan(scan_id, out_dir, file_types)elif 'minos' not in args.task_data and args.id == 'ALL' or args.id == 'all':  # download entire releaseif len(file_types) == len(FILETYPES):print('WARNING: You are downloading the entire MP release which requires ' + RELEASE_SIZE + ' of space.')else:print('WARNING: You are downloading all MP scans of type ' + file_types[0])print('Note that existing scan directories will be skipped. Delete partially downloaded directories to re-download.')print('***')print('Press any key to continue, or CTRL-C to exit.')# key = raw_input('')key = input('')out_dir = os.path.join(args.out_dir, RELEASE)download_release(release_scans, out_dir, file_types)if __name__ == "__main__": main()

创建一个文件夹，用于存放数据集，比如data

下载所以MP3D的数据，1.3T左右，执行命令：

python download_mp.py --task habitat -o ./data

打印信息：

Downloading MP task data for ['habitat'] ...
Done downloading task_data for ['habitat']
Press any key to continue on to main dataset download, or CTRL-C to exit.
WARNING: You are downloading the entire MP release which requires 1.3TB of space.
Note that existing scan directories will be skipped. Delete partially downloaded directories to re-download.
***
Press any key to continue, or CTRL-C to exit.

Downloading MP release to ./data/v1/scans...

Downloading MP scan 1pXnuDYAj8r ...
下载完成：./data/v1/scans/1pXnuDYAj8r/cameras.zip
开始下载 ./data/v1/scans/1pXnuDYAj8r/matterport_camera_intrinsics.zip (0.10 MB)
下载进度: 100% (0.10/0.10 MB)
下载完成：./data/v1/scans/1pXnuDYAj8r/matterport_camera_intrinsics.zip
开始下载 ./data/v1/scans/1pXnuDYAj8r/matterport_camera_poses.zip (0.61 MB)
下载进度: 100% (0.61/0.61 MB)
下载完成：./data/v1/scans/1pXnuDYAj8r/matterport_camera_poses.zip
开始下载 ./data/v1/scans/1pXnuDYAj8r/matterport_color_images.zip (537.00 MB)
下载进度: 100% (537.00/537.00 MB)
............
下载完成：./data/v1/scans/1pXnuDYAj8r/undistorted_color_images.zip
开始下载 ./data/v1/scans/1pXnuDYAj8r/undistorted_depth_images.zip (1282.63 MB)
下载进度: 100% (1282.63/1282.63 MB)
下载完成：./data/v1/scans/1pXnuDYAj8r/undistorted_depth_images.zip
开始下载 ./data/v1/scans/1pXnuDYAj8r/undistorted_normal_images.zip (2808.76 MB)
下载进度:  90% (2528.00/2808.76 MB)
 

下载过程中，场景数据的文件：

17DRP5sb8fy任务包含的内容：

...............

任务数据文件：

指定某个场景ID，进行下载，执行命令：

python download_mp.py --task habitat -o ./data --id 17DRP5sb8fy

打印信息：

Downloading MP scan 17DRP5sb8fy ...
开始下载 ./data/v1/scans/17DRP5sb8fy/cameras.zip (0.00 MB)
下载进度: 100% (0.00/0.00 MB)
下载完成：./data/v1/scans/17DRP5sb8fy/cameras.zip
开始下载 ./data/v1/scans/17DRP5sb8fy/matterport_camera_intrinsics.zip (0.05 MB)
下载进度: 100% (0.05/0.05 MB)
........
下载完成：./data/v1/scans/17DRP5sb8fy/image_overlap_data.zip
开始下载 ./data/v1/scans/17DRP5sb8fy/poisson_meshes.zip (136.50 MB)
下载进度: 100% (136.50/136.50 MB)
下载完成：./data/v1/scans/17DRP5sb8fy/poisson_meshes.zip
开始下载 ./data/v1/scans/17DRP5sb8fy/sens.zip (1402.69 MB)
下载进度: 100% (1402.69/1402.69 MB)
下载完成：./data/v1/scans/17DRP5sb8fy/sens.zip
Downloaded scan 17DRP5sb8fy

分享完成～

http://www.dtcms.com/a/291682.html

相关文章：

  • Aspose.Cells 应用案例：法国能源企业实现能源数据报告Excel自动化
  • UE创作一个可以变更列数的万能表格
  • Excel file format cannot be determined, you must specify an engine manually.
  • 如何撤销Git提交误操作
  • 实用资源分享：50款出入库单据Excel模板集合
  • DFS习题篇【下】
  • 北京养老金计算公式网页实现案例：从需求分析到架构设计
  • 业务流逻辑如何搭建？为何橙武平台选用了 LogicFlow？
  • 【MyBatisPlus】一文讲清 MyBatisPlus 基本原理及基本使用方式
  • EMA《2025-2028年药品监管中的数据与AI 1.3版》信息分析
  • 深度分析：Kimi K2开源模型
  • 拆分、合并PDF
  • Qt基本控件使用：按钮、标签、文本框等
  • docker阿里云安装
  • [2025CVPR]ViKIENet：通过虚拟密钥实例增强网络实现高效的 3D 对象检测
  • AI Agent-Manus 构建经验解读(下)
  • powerquery如何实现表的拼接主键
  • mybatis多对一一对多的关联及拼接操作以及缓存处理
  • Java 与 Android 回收机制深度解析
  • 行业出海研究报告
  • Apache Ignite 中的 SQL 模式（Schema）管理机制
  • Qt字符串处理与正则表达式应用
  • MCP vs 传统集成方案：REST API、GraphQL、gRPC的终极对比
  • 使用vue-pdf-embed发现某些文件不显示内容
  • Jenkins接口自动化测试（构建）平台搭建
  • Jenkins 多架构并发构建实战
  • 计算机网络：连接世界的数字脉络
  • Python爬虫实战：研究pymorphy2库相关技术
  • JVM：工具
  • 字节跳动视觉算法面试30问全景精解