dataset-buiding-projects

---

Dataset: Tartu building project register

(19th-century urban construction permits and plans)

---

1. Access & allowed use

This dataset is provided for:

• ✔ hackathon research and prototyping
• ✔ visualizations and models
• ✔ presentations and demo publications

Not allowed:

• ✘ commercial reuse
• ✘ re-publishing raw data outside this repository
• ✘ identifying private individuals beyond historical context

Citation requirement

National Archives of Estonia. Tartu Architectural Projects dataset, 2026.

---

2. Data structure

Main file

data/raw/*.csv

• Delimiter: |
• Encoding: UTF-8

---

Row represents

One building project record (permit / plan entry)

Not:

• a single building
• a finished structure
• a full property history

---

Key fields

Field	Meaning	Notes
id	unique identifier	stable
Projekti kuupäev	project date	administrative date
Linn / Tänav / Maja nr	location	historical naming
Otstarve	intended use	planned function
Ehituse tüüp	project type	e.g. new construction / extension
Välisseina materjal	wall material	planned material
Korruseid	floors	planned structure
Omanikud	owners / commissioners	JSON-encoded persons
Failid	linked drawings	archival references

---

Extended fields (full schema)

The dataset contains a large number of fields describing archival metadata, building characteristics, and project details. They are grouped below for clarity.

---

📁 Archival reference

Field	Meaning
Fondi nimi	Archive collection name
Fond	Collection identifier
Nimistu	Inventory number
Säilik	File / unit
Lehekülg	Page
refcode	Archival reference code
Pages Sequence	Page order within file

---

🖼️ Linked materials

Field	Meaning
Failid	Linked image/drawing references
Krundiplaan	Site plan
Ehitusplaan vanas / uues	Building plan (before / after)
Korruseplaan vanas / uues	Floor plan
Lõige vanas / uues	Section drawing
Tänavapoolne fassaad vanas / uues	Street-facing facade
Fassaad2–4 vanas / uues	Additional facades
Sisustusplaan vanas / uues	Interior plan

---

📍 Location

Field	Meaning
Linn	City
Linnaosa	District
Tänav	Street (historical)
Tänav (uus)	Street (new=1940)
Maja nr. (uus)	New (=1940) house number
Kinnistu nr.	Property number

---

🏗️ Building description

Field	Meaning
Otstarve	Intended use
Ehituse tüüp	Type of construction
Kapitaalsuse grupp	Building category / durability class
Välisseina materjal	Outer wall material
Vaheseina materjal	Inner wall material
Viimistlus	Finishing
Katusematerjal	Roof material

---

⚙️ Infrastructure & amenities

Field	Meaning
Küte	Heating
Vesi	Water supply
Soe vesi vanas / uues	Hot water
Vann vanas / uues	Bath
Kuivkäimla	Dry toilet

---

🏢 Structure (before vs after)

Field	Meaning
Korruseid vanas / uues	Number of floors
Soklikorrus vanas / uues	Basement level
Mansardkorrus vanas / uues	Attic floor
Solk vanas / uues	Sewage / drainage

---

🏠 Apartment structure

Field	Meaning
1–10-toalisi kortereid vanas / uues	Number of apartments by size
rohkem kui 10-toalisi kortereid vanas / uues	Large apartments

(Note: some fields may appear duplicated or inconsistently named — this reflects the source structure.)

---

👥 Ownership

Field	Meaning
Projekti tellija \ Omanikud	Project commissioner / owner(s)

---

📅 Project metadata

Field	Meaning
Projekti nr.	Project number
Projekti kuupäev	Project date
Projekti kuupäeva täpsus	Date precision
Projekti kinnitamise kuupäev	Approval date
Projekti kinnitamise kuupäeva täpsus	Approval date precision

---

📝 Additional notes

Field	Meaning
Märkused	Free-text notes

---

⚠️ Important notes about the schema

• Many fields exist in pairs (vanas / uues) → representing before vs planned state, not time series

• Not all fields are filled for all records → absence often reflects administrative relevance

• Some column names are inconsistent or duplicated → this reflects the historical database export, not errors

---

Why this structure matters

The dataset contains:

• archival metadata
• spatial data
• building descriptions
• transformation data (before → after)
• linked visual materials

---

About the owner field

The Omanikud field contains structured JSON embedded in text, for example:

[{"Eesnimi": "...", "Perenimi": "..."}]

This allows reconstruction of:

• ownership networks
• commissioners vs builders
• institutions vs individuals

This is structured data embedded in a string.

---

About materials and structure fields

These describe the planned building, not necessarily the built one.

They reflect:

• regulatory expectations
• reporting practices
• administrative requirements

Do not interpret them as a direct description of the historical cityscape.

3. Associated image materials (temporary)

Some project records have linked visual materials:

They will be provided separately during the hackathon only and will be available via link through this README.

These files are not part of the persistent dataset and should be treated as a temporary research environment.

Relationship to the CSV:

one record → zero, one, or multiple images images correspond via identifiers in the Failid field absence of an image does not mean absence of a project

⚠ Important technical note

You should design your workflow so that results do not depend on keeping the image files locally after the hackathon. Extract features, annotations, or measurements instead of building pipelines that require long-term access to the images.

---

🔒 Temporary access to image materials

A ZIP archive with digitized project drawings is provided.

These files are provided only for the duration of the hackathon as temporary research material.

---

Allowed during the event

• ✔ viewing the images
• ✔ analysing visual features
• ✔ extracting measurements or annotations
• ✔ using derived data (coordinates, counts, classifications, embeddings, notes)
• ✔ including small illustrative excerpts in presentations

---

Not allowed

• ✘ storing the images permanently on personal devices
• ✘ uploading images to external services
• ✘ training reusable datasets containing the images
• ✘ redistributing the images after the event
• ✘ keeping local copies after the hackathon ends

---

After the hackathon, participants must delete all local copies of the image files.

---

Derived results (allowed to keep)

• measurements
• annotations
• model outputs
• statistics
• visualizations
• manually redrawn diagrams

---

Why this restriction exists

The images are shared under a limited access agreement with the holding institution. We can enable research access, but not public redistribution.

If you are unsure whether something counts as derived data, please ask in the Issues tab before using it.