AI Infrasolutions panorama

Session-based panorama deliveries. Each session folder at the bucket root becomes one set, named after the folder.

Every row in the CSV must resolve to a real photo — there is no allowance for missing ones. See Tolerance before your first delivery.

Folder layout#

<dataset bucket>/
└── 2025-07-04_vrv-64-n_pr25-bnv-001_<uuid>/   session (one set)
    └── panorama_1/
        ├── Recording.csv                      exact name, exact case, tab-separated
        ├── Recording_20250704_065437/
        │   └── 04428.jpeg                     Filename column + `.jpeg`
        └── Recording_…

Any folder at the bucket root that directly contains panorama_1/ is one of the sessions — recognised by its contents, not its name.

A date-prefixed folder name is the usual convention, not a requirement: any root folder that directly contains panorama_1/ is a session. A delivery with no such folder fails.

Photos live in subfolders of panorama_1/. The subfolder name comes from the CSV's Directory column, so the two must agree exactly.

Every delivered photo is opened and checked, not just listed:

Every delivered photo must start with a JPEG marker and end with an end-of-image marker. This catches a file that is not a JPEG whatever its extension says, and the truncated upload — which a size check cannot see, because a truncated file just looks smaller and nothing knows how big it should have been.

Recording.csv#


Separator	Tab
Preamble	None — the header row comes first
Columns	Addressed by name, case-insensitively. Order does not matter
Extra columns	Unnamed trailing columns are tolerated
Decimals	Both `.` and `,` accepted
Photo names	Joined from the `Directory` and `Filename` columns, then `.jpeg` appended — case-sensitive
Which file	`Recording.csv` directly inside `panorama_1/` (exact name, exact case)

Column	Type	Meaning
`Directory`	text	Photo's subfolder, relative to `panorama_1/` — joined with `Filename` to form the path
`Filename`	text	Photo name without extension — `.jpeg` is appended; never `.jpg`
`X/Long`	number	Easting, in the dataset's CRS
`Y/Lat`	number	Northing
`Z`	number	Height
`Pan`	number	Heading, degrees
`Tilt`	number	Degrees
`Roll`	number	Degrees
`Timestamp`	timestamp	GPS Time of capture

Columns are addressed by name, so order does not matter — but the nine names above must be there. Extra trailing cells (unnamed accuracy columns) are ignored.

Do not re-save the file as comma-separated. A comma-separated file is not read as a slightly different export; every line becomes a single field and the delivery fails on missing columns.

Filename carries no extension. The resolved path is <Directory>/<Filename>.jpeg, and the extension is always .jpeg — never .jpg. See Filenames.

Timestamps#

Timestamp is GPS Time seconds since 1980-01-06 (e.g. 1435647298.756574 → 2025-07-04T06:54:40.757Z).

Deliver the recorder's GPS Time seconds. Do not convert to Unix epoch first.

Example#

Directory	Filename	X/Long	Y/Lat	Z	Pan	Tilt	Roll	Timestamp
Recording_20250704_065437	04428	177394.99911378	471544.14159442	49.778086	8.983611	-1.645189	-0.093183	1435647298.756574

Tolerance#

Every CSV row must resolve to a delivered photo. One unresolvable row fails the run — but every row is checked first, so one pass reports them all.
Photos in the delivery that the CSV does not reference are not indexed (noted, not an error).

Every row is still processed and reported before the run exits, so a single pass tells you everything that is wrong rather than one problem at a time.

Practically: export the CSV from the same source that produced the folder, and don't hand-edit either afterwards.

Coordinate reference system#

Nothing is reprojected. X, Y and Z are taken as delivered and stamped with the dataset's 3D SRID. See Coordinates are not reprojected.

Pan is a compass heading (north is 0°, clockwise); Tilt and Roll follow. See Rotation angles.

Ingest parameters#

Key	Required	Type	Meaning
`anonymisation`	yes	string — recognised: `already_anonymised`, `auto_anonymise`	Compared exactly: `already_anonymised` marks the photos as already blurred, `auto_anonymise` queues them for blurring. There is no Ignore — these are street panoramas.
`height_above_ground`	yes	number	Camera height above the road surface, in metres.

What you get back#

Indexed panorama records with positions and orientations, grouped by session.
A tilepack (.tpc) per photo. Omnibase tenants view the JPEG; DZP is not generated at ingest.
Blurred derivatives, if anonymisation is auto_anonymise.

Common failures#

Symptom	Cause
"no session folders found" / "the delivery has no sessions"	Nothing at the bucket root contains a `panorama_1/` folder
"no `Recording.csv` directly inside …"	The file is missing, spelled differently, or sits in the session root instead of `panorama_1/`
Every row reports a missing photo	The CSV is comma-separated rather than tab-separated, or `Filename` includes the extension
Fewer photos indexed than delivered	Photos on disk that the CSV doesn't list are skipped — add rows for them, or remove them
Capture times decades out	`Timestamp` was delivered as Unix epoch instead of GPS seconds