Hugging Face's logo

Datasets:
fengmap-official
/
indoor-semantic-sample

Modalities:
Image
Tabular
Text
Formats:
parquet
Languages:
English
Chinese
Size:
< 1K
Tags:
geospatial
image
indoor-maps
spatial-semantics
navigation
robotics
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
indoor-semantic-sample / README.md
fengmap's picture
fengmap
Translate dataset card to English
e7550e3 verified
preview code
|
Raw
History Blame Contribute Delete
6.06 kB
metadata 
pretty_name: Fengmap Indoor Semantic Map Sample Dataset
language:
  - en
  - zh
license: other
license_name: fengmap-non-commercial-dataset-terms-v1.0
license_link: >-
  https://huggingface.co/datasets/fengmap-official/indoor-semantic-sample/blob/main/LICENSE
size_categories:
  - n<1K
tags:
  - geospatial
  - image
  - indoor-maps
  - spatial-semantics
  - navigation
  - robotics
configs:
  - config_name: floors
    default: true
    data_files:
      - split: train
        path: data/floors.parquet
  - config_name: maps
    data_files:
      - split: train
        path: data/maps.parquet

Fengmap Indoor Semantic Map Sample Dataset

Dataset version: v1.0
Semantic format specification version: v0.2
Release date: August 14, 2026
Dataset size: 5 indoor maps across 37 floors
Permitted use: Non-commercial learning, research, education, and technical validation only

Dataset Overview

The Fengmap Indoor Semantic Map Sample Dataset is a public test dataset designed for indoor spatial understanding, spatial relationship analysis, map SDK integration, and navigation application validation.

Our goal is to provide a practical public sample that can be parsed, run, and validated, so discussions around indoor map semantics can extend beyond theory. Developers can use real-world indoor spatial data to validate spatial entities, relationships, navigation topology, and the feasibility of combining indoor maps with AI, robotics, and related applications.

This dataset was collected independently by Fengmap and subsequently organized, processed, and semantically annotated. Public materials identify test environments by map ID; the names of the shopping malls are not disclosed.

Published Contents

This Hugging Face Dataset repository provides:

  • Floor map screenshots for 5 maps across 37 floors;
  • Parquet views containing map summaries and floor screenshots for the Dataset Viewer;
  • The official v1.0 ZIP archive, identical to the GitHub Release asset;
  • The Semantic Output Specification v0.2 document;
  • Custom non-commercial terms and file checksums.

The official ZIP archive contains standard semantic data and floor screenshots for all five maps. The semantic data covers entities, relationships, navigation nodes, and navigation edges.

Map ID Floors Entities Relationships Navigation Nodes Navigation Edges Semantic Data Size
123456 10 4,222 5,130 2,307 5,568 5.84 MiB
272862 6 3,876 4,251 1,243 2,792 4.60 MiB
709246 7 3,190 4,162 1,730 4,367 5.06 MiB
910807 7 6,700 7,730 3,058 6,951 8.33 MiB
911907 7 2,648 3,047 1,437 3,426 3.48 MiB
Total 37 20,636 24,320 9,775 23,104 27.32 MiB

Repository Structure 

.
├── README.md
├── LICENSE
├── checksums.sha256
├── data/
│   ├── maps.parquet
│   ├── floors.parquet
│   └── floor-images/
│       └── {mapId}/{floorName}.png
├── docs/
│   └── semantic specification document
└── release/
    └── fengmap-semantic-sample-dataset-v1.0.zip

maps.parquet contains summary statistics for the five maps. floors.parquet contains 37 floor records with embedded screenshots for the Dataset Viewer. The Parquet files are derived views intended for convenient browsing and analysis. The ZIP archive under release/ is the authoritative distribution package.

Usage

Load and browse the floor screenshots with Hugging Face Datasets:

from datasets import load_dataset

floors = load_dataset(
    "fengmap-official/indoor-semantic-sample",
    "floors",
    split="train",
)

maps = load_dataset(
    "fengmap-official/indoor-semantic-sample",
    "maps",
    split="train",
)

Download the official ZIP archive:

hf download fengmap-official/indoor-semantic-sample \
  release/fengmap-semantic-sample-dataset-v1.0.zip \
  --repo-type dataset \
  --local-dir .

ZIP SHA-256:

17208aa684dd357b5dfe4e8717a08d27d6df3a4fbbb65a025cbcb6f9dde0cd2c

See checksums.sha256 for checksums of the primary release files.

Documentation and Demos

Permitted Use and Restrictions

This test dataset may be used only for non-commercial learning, research, education, public-interest activities, and technical validation. See LICENSE for the complete terms.

This dataset is not a production map service and must not be used as the sole source of information for live shopping mall operations, fire evacuation, emergency response, or any other safety-critical system. Indoor spaces, floor layouts, store names, facility status, and access conditions may change over time.

Without authorization, users must not use layouts, names, brand combinations, or other information in this dataset to identify, disclose, or distribute the identity of any undisclosed shopping mall. Third-party brand identifiers must not be used for advertising, endorsement, or any purpose unrelated to this test dataset.

Citation 

@dataset{fengmap_indoor_semantic_sample_2026,
  author    = {Fengmap},
  title     = {Fengmap Indoor Semantic Map Sample Dataset},
  year      = {2026},
  version   = {1.0},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/fengmap-official/indoor-semantic-sample}
}

Contact