Private dataset workspace
Datum LabelHub
Upload façade folders, label the building envelope and every opening, then export clean training data.
Private dataset workspace
Upload façade folders, label the building envelope and every opening, then export clean training data.
Training data, kept understandable
Create one dataset per job, source, or photo folder. Progress, opening counts and exports stay together.
Approve or deny
Check front-of-house photos before they enter a labelling folder. Source details stay attached forever.
Loading datasets…
Elevation improvement loop
Select one or more folders below. Complete usable labels become permanent tests; drafts stay visible but cannot enter training data.
Then upload individual images or a whole folder. Folder names are preserved in the export.
Source review
Only approved photos enter a labelling folder. Every decision keeps its source record.
Runs on the server
Choose an area and how many new unique houses you want. The job keeps running here, skips listings and pixels already known, and puts compact images directly into Pending review.
The server will continue even if you leave this screen.
Drag this button to your bookmarks bar. On the Zoopla London houses results page, click it, choose 1–4 pages, and it will collect each listing's first photo into a small source file.
Collect Zoopla London pageAdd sourced front-view photos when you are ready.
Dataset
—
Open an image to review or continue its labels.
Add individual photos or upload a complete folder.