Dimensioning and Weighing Station Records

The buyer's problem

A robot learning team that handles cartons, polybags and loose items needs to know how big and heavy they really are. Product master data supplies nominal values. Stations that measure each unit capture what actually arrived, including crushed corners, bulging bags and mislabeled packs. The gap between the two is information.

Perception engineers use measured dimensions to validate size estimation from depth. Planning teams use weights to set payload limits and to check whether a picked unit matches the expected one. Researchers want a ground-truth channel for images they already hold. This page covers what such records contain, how they differ from master data, and what to ask for.

Nominal data and measured data

GS1 publishes the Package and Product Measurement standard, a global standard for the nominal measurement attributes of retail and non-retail products from consumer unit to pallet. Its current release, 3.3, was ratified in August 2026, and it replaced the earlier GDSN Package Measurement Rules (GS1). GDSN lets trading counterparties share product master data through GS1-certified data pools and the GS1 Global Registry (GS1).

The key word is nominal. Master data states what a product is supposed to measure and weigh. A station reading states what one physical unit measured, under one device, at one moment. When you buy records, ask which kind you are getting, and whether the supplier can supply both and the difference between them. For the catalog side, see product catalog data for robot perception and SKU pack instructions.

What this data can support, and what it cannot

Can support. Ground-truth size and mass labels for images taken at or near the station; checks on depth-based size estimation; payload and fit modeling; detection of deviations such as overweight or oversized units; and correction of nominal values used by a simulation or planner. When station cameras exist, a record plus an image gives a labeled sample without annotators.

Cannot support. A dimension is a bounding box. It does not describe shape, surface, rigidity, center of mass or how an item behaves when lifted. Weight is one scalar and not a force profile. Records have no visual information by themselves, and no robot actions. For handling you still need demonstrations; see the distinction between recognition and manipulation data. Records often contain customer and employee identifiers, and parcel data can refer to recipients.

Typical fields

Logistic units can be keyed. The GS1 Serial Shipping Container Code is an 18-digit key that uniquely identifies a logistic unit such as a case, pallet or parcel (GS1); where a site uses it, the key can join a station reading to shipping data.

Illustrative record set

# Illustrative only - synthetic values
records: dim_weigh_station_v0
site_id: site_063
device: inline_dimensioner_A (type stated), conveyor_scale_B
record:
  unit_id: u_0004417
  measured: {l_mm: 312, w_mm: 214, h_mm: 98, weight_g: 1045}
  nominal_from_master: {l_mm: 300, w_mm: 210, h_mm: 100, weight_g: 1000}
  t_utc: 2026-03-04T14:22:11Z
  status: valid
  image_ref: img_88231 (optional)
notes: orientation convention, units, rounding stated in data dictionary
excluded: [customer_name, address, employee_id]

Quality and acceptance pointers

• Units and conventions. Millimeters or inches, longest-side rule, orientation. A data dictionary should state these.
• Device accuracy. Ask for the device type, calibration record and stated resolution. Do not assume parity across sites.
• Valid vs overridden readings. A high override share signals trouble.
• Pairing with images. If images exist, measure the offset between record time and image time.
• Coverage. Variety of item types, bags vs rigid cartons, and size range. A set of uniform cartons teaches little about soft goods.
• Master-data comparison. Check the share of units where measured and nominal values disagree, and whether the supplier can release both.

Supplier types

Parcel carriers' hubs, fulfillment centers and 3PLs run dimensioners and checkweighers. Measurements may be stored in WMS, TMS or carrier billing systems. Because 3PLs measure client goods, rights typically sit with the client; see client-owned inventory rights, 3PL providers as warehouse data suppliers and receiving and inbound checks.

Limitations and additional preparation

Structured records alone are not training data for a humanoid. They need time alignment, a join key, removal or tokenizing of customer and employee identifiers, and an explicit statement of measurement method. Clipped or capped values at the device limit, parcels measured twice and units re-measured after repack can distort statistics. The records describe items at a station, not items in a bin or on a shelf. Combine them with images and with demonstrations from the target hardware, and hold out evaluation by site and item type.

Using measured records with images

The most direct use of a dimensioning record is as a label for an image taken at the same moment. If a station camera captured the parcel, the record gives size and mass for free, and a team can train or test size estimation from depth or monocular cues. Check that the image and record refer to the same unit: some conveyors have two or three units between the sensor and the camera.

A second use is to correct nominal data in simulation. Replace master-data dimensions with measured distributions and the simulated boxes behave more like the real ones. A third is to bound payload: the heaviest observed unit in each class sets a realistic limit, and the distribution shows how often it occurs. Records can also support evaluation, for example by defining a test set of units whose measured size differs from nominal. For a related use of system records as outcome labels, see pick confirmation as automatic labels.

Questions to ask suppliers

1. Which devices measure dimensions and weight, and what are their stated resolutions?
2. How often are devices calibrated, and are calibration logs available?
3. Are nominal master-data values available alongside measured values?
4. How are readings keyed to items, and does the key match any image or event data?
5. What share of readings are manual overrides or retries?
6. Are orientation and measurement conventions documented?
7. Do records include recipient, customer or employee identifiers, and can they be removed or tokenized?
8. Does the supplier own the data, or does a client or carrier?
9. Are images captured at the station, and with what timing relative to the reading?

Discuss your requirements

If you can state the item types, fields and pairing you need, describe the requirement to SourceX. SourceX reviews it, follows up about potential sources among US businesses, checks the data and the supplier's licensing permissions, and coordinates the license and delivery. A request does not guarantee a match. See the humanoid training data guide and the warehouse and fulfillment hub, and for catalog context the supply chain and logistics data page.

Discuss your humanoid training-data requirements with SourceX

Sources

• GS1 Package and Product Measurement Standard (formerly GDSN Package Measurement Rules) — nominal measurement standard and GDSN data sharing.
• Serial Shipping Container Code (SSCC) — logistic unit key.