Short answer
Pre-use and pre-delivery inspection records pair a structured checklist, with pass, fail and note fields, with photos of the equipment at a known moment. They can support condition recognition, checklist-step grounding and inspection evaluation sets. They contain no robot actions, so they cannot train an inspecting humanoid to move, reach or position a camera.
Why the pairing matters
A photo alone shows a surface. A checklist alone shows a verdict. Together they show what a trained person looked at, what they decided, and what the equipment looked like when they decided it. That link is what inspection models usually lack when a buyer licenses a pile of unlabeled images.
Fleets, equipment rental companies and dealers generate these records as a side effect of running assets. A driver walks around a vehicle before a shift. A rental yard checks a machine before it leaves and again when it returns. A dealer documents condition before handing a unit to a customer. Each event can leave a form, a set of photos and a signature.
This page covers how to source those records. For the broader question of linking images to pass, fail and rework outcomes, see inspection images linked to outcomes. The overall map of the section is in the humanoid training data guide and the inspection, defect and failure data hub.
What a checklist-and-photo record can contain
Records differ widely by operator, so treat the following as things to confirm rather than expect:
- Checklist items. Fixed lists such as tires, lights, fluid leaks, guards, hoses, labels and fasteners, each with a result. Some systems store only exceptions.
- Photos. Mandatory shots at fixed positions, or free-form photos attached to a failed item. Ask which one applies, because free-form photos are harder to align to checklist items.
- Context. Asset ID, odometer or hour meter, location, time, weather and inspector role.
- Outcome. Released, released with note, held for repair, or returned damaged, plus any later work order.
- Comparison pairs. Pre-delivery and return photos of the same asset, which show changes over time.
Organizations that run a quality management system often keep documented information about checks because standards of that kind ask for it. ISO 9001 describes requirements for documented information inside a generic framework (ISO, 2026). That tells you why records may exist; it says nothing about whether a given supplier's checklists are complete or usable as training data.
What the data can support
- Condition recognition. Classifying tires, hoses, guards or labels as acceptable or not acceptable when the photo and the verdict agree.
- Checklist grounding. Linking an item's wording to the image region it refers to, if annotators or inspectors marked the area.
- Change detection. Comparing outbound and return photos of the same asset.
- Inspection language. Short free-text notes such as "weeping seal, rear left" paired with an image.
- Evaluation sets. Held-out assets or sites to test whether a model's verdicts match human inspectors.
A buyer building an inspecting robot can use such data to evaluate perception, but should read active inspection by humanoids for what the robot itself still needs.
Limits: what this data cannot support
Control. A humanoid that performs a walkaround needs to plan a path, stop at the right standoff, crouch or reach, and aim a sensor. Checklist photos contain no body motion, no joint states, no commanded actions and no timing of how the inspector moved. Control training needs robot or teleoperation data; photos and forms alone do not provide it.
Viewpoint. Photos are taken from a human's hand-held phone or tablet at human heights and angles. A robot's camera height, lens and approach angle may differ. Check this early with imaging conditions metadata.
Label noise. A "pass" on a busy yard may mean "looked at it quickly." Photos taken only for failures bias the set toward defects and hide the normal appearance. Unfailed items without photos give no negative examples.
Coverage. One fleet type, one climate or one rental category may not transfer to others. Hidden areas such as undersides or inside engine bays may never be photographed.
Illustrative record schema
The following is an illustrative schema with synthetic values, not an existing dataset. Adapt the fields to your task.
# Illustrative only: synthetic values, not a real or available dataset
record_id: rec_00041
asset: {asset_id: asset_A17, category: "compact loader", site_id: site_003}
event: {type: pre_delivery, timestamp: "2026-01-14T08:20:00Z", inspector_role: yard_tech_B2}
checklist:
- item_id: hyd_hoses
result: fail
note: "abrasion near fitting, left boom"
photo_ids: [img_0007, img_0008]
- item_id: tracks
result: pass
photo_ids: [img_0011]
photos:
- photo_id: img_0007
device: handheld_phone
shot_position: front_left_quarter
resolution: 4032x3024
exif_present: true
outcome: {disposition: held_for_repair, work_order_ref: wo_5521}
rights: {license_scope: "training and evaluation", redistribution: no}
Quality and acceptance criteria
Set acceptance thresholds yourself, based on the use. Examples a buyer might choose to test on a sample of records:
- Every failed item has at least one photo that shows the stated problem on inspection by a reviewer.
- Photo-to-item mapping is explicit, not inferred from filename order.
- The same asset appears with at least two events if change detection matters. The number depends on your design.
- Image metadata is retained or its removal is documented, because lighting and device information help later analysis.
- Disposition and any follow-up work order can be joined where you need outcomes; see equipment wear and damage images for related label design.
Hold out entire sites or asset classes for evaluation, not random images, so near-duplicate photos of one machine do not leak across splits.
Supplier types and what each tends to hold
Fleet operators often hold high-frequency, short checklists with a few photos. Equipment rental companies may hold paired outbound and return records. Dealers may hold condition documentation around sale or handover. See equipment rental inspection and repair for the repair side, and preventive maintenance checklists for scheduled-service lists that differ from walkaround checks.
Permissions and provenance
Photos can include license plates, serial plates, yard signage, customer names on documents, and people in the background. Checklists carry inspector names or initials, and they may carry customer identifiers. Ask who owns the records, whether customers or drivers were notified, and whether a lessor, manufacturer or customer holds rights in the content. Blurring identifiers reduces exposure but does not by itself settle permission questions. This is not legal advice.
Questions to ask suppliers
- Are photos mandatory at fixed positions, or only attached to failed items?
- Is each photo tied to a specific checklist item by ID?
- What devices took the photos, and is original metadata kept?
- Do records include outbound and return events for the same asset?
- How are disposition and follow-up work orders linked?
- Who completed the checklists, and how were they trained to judge pass and fail?
- Have checklist templates changed over time, and can versions be identified?
- Which records contain customer names, plates or people, and who can license them?
- Can you provide a sample of records, including failed items, before a license is agreed?
Discuss your requirements
If you need checklist-and-photo records, describe the asset types, modalities and permitted uses to SourceX. SourceX looks for US businesses that hold suitable material, checks the data and the supplier's licensing permissions, and coordinates the license and delivery. A request does not guarantee a match; SourceX follows up about potential sources. The catalog pages on images and inspection photos and inspection reports describe related data kinds.
Explore sourcing and licensing options with SourceX
Sources
- ISO 9001:2026 Quality management systems - Requirements — documented information within a generic quality management framework.