Back to datasets
Dataset assetOpen Source CommunityIndoor PositioningDynamic Object Recognition

NAVER LABS Indoor dataset

These datasets were obtained in a real department store environment and contain hundreds of dynamic objects, such as humans. Please refer to the NAVER LABS website for detailed explanations.

Source
github
Created
Nov 18, 2021
Updated
Nov 19, 2021
Signals
103 views
Availability
Linked source ready
Overview

Dataset description and usage context

NAVER LABS Indoor dataset - LiDAR API

Dataset Description

  • Environment Type: Complex indoor environment located in a real department store.
  • Dynamic Objects: Contains hundreds of dynamic objects, such as humans.
  • Data Download: The dataset can be downloaded via this link, but note that requests for the HD Map & Localization Dataset are limited to researchers and organizations in Korea.

Data Format

  • File Structure:

    data_path (/nvlabs/abs_dir in launch/lidar_publisher.launch file) ├── images ├── pointclouds_data ├── camera_parameters.txt ├── groundtruth.hdf5 └── map.pcd

  • Data Loading Example: cpp // 加载第i个点云数据 pcl::PointCloud::Ptr srcCloud(new pcl::PointCloud); *srcCloud = *loader.cloud(i);

    // 加载第i个位姿 Eigen::Matrix4f pose = loader.pose(i);

How to Run

  • Build and Launch: bash $ catkin build naverlabs_api $ roslaunch naverlabs_api lidar_publisher.launch

  • LiDAR Mode Selection: Users can choose among lidar0, lidar1, and both modes.

Applications

  • Dynamic Object Processing: The ERASOR tool can be used to remove traces of dynamic objects.
  • Effect Comparison:
    • Before processing: Traces of dynamic objects are visible.
    • After processing: Traces of dynamic objects are removed.
Need downstream help?

Pair the dataset with AI analysis and content workflows.

Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.

Explore AI studio