Short answer
Equipment wear and damage images come from maintenance inspections, repair photos and condition surveys, and show corrosion, abrasion, cracks, leaks, deformation and missing parts on real equipment. They support condition recognition and report generation when linked to asset records and a repair outcome. They do not show how to repair or handle the part, and photo quality varies widely.
What a buyer is really after
Condition imagery answers a perception question: what state is this component in, and how does that state compare with the last visit? Systems that walk a facility, read a gauge or check a pump housing need to recognize worn belts, weeping seals, rust blooms and loose fasteners in normal working conditions, not in a lab.
Many useful photos may already exist inside maintenance organizations. Technicians may attach pictures to work orders, auditors photograph findings, insurers' surveyors document damage, and fleet operators image components at teardown. Ask how much exists and where it sits; such material is typically scattered and was collected to justify a repair or to document a finding, not to train a model.
This page covers sourcing such imagery. The manufacturing-line equivalent, with parts and defect codes, is in sourcing defect image datasets. Imagery of buildings and civil assets sits in building and infrastructure defects.
Where condition photos live
| Supplier type | Typical material | Strength | Common weakness |
|---|---|---|---|
| Facilities and building maintenance | Photos attached to work orders | Variety of equipment and wear | Inconsistent framing and lighting |
| Fleet and vehicle service | Teardown and warranty photos | Component-level detail | Close crops without context |
| Industrial maintenance teams | Condition-monitoring rounds | Repeat visits to the same asset | Dirty lenses, mixed devices |
| Inspection and survey firms | Report images with findings | Written finding for each image | Rights often held by their client |
| Equipment rental and resale | Return-condition photos | Consistent angles and checklists | Cosmetic damage dominates |
| Insurers and adjusters | Damage documentation | Clear damage framing | Heavy personal data and claim context |
Each supplier type has a different rights position, which is why work order histories are often a practical starting point: the photo may already be attached to a record with a symptom, diagnosis and fix.
What the images can support, and what they cannot
They can support:
- Condition classification and detection of visible wear states, damage types and missing components.
- Change detection, when the same asset is photographed at several visits.
- Report generation and image-text pretraining, when each image has a technician's description. Maintenance logbook text from aviation, automotive and facilities domains has been released as research resources (Akhbardeh et al., 2020), which shows free-text condition notes are kept in these fields and can be paired with images when the records are linked. See maintenance logbook text for the text side.
- Triage evaluation, such as whether a model correctly escalates a damaged part.
They cannot support:
- Manipulation or repair skill. A photograph does not contain the motion, force or tool use that fixes the fault. For that, a buyer needs repair demonstrations; see repair video demonstrations.
- Wear progression, severity or remaining life, without measurements, operating hours or later outcomes.
- Coverage of the working environment. Close crops taken for a report rarely show how an inspecting robot would see the asset from a fixed height or approach path.
- Ground-truth causes. A photo of a worn bearing does not establish why it wore.
Illustrative package
Counts and fields are placeholders the buyer sets.
| Component | Illustrative content |
|---|---|
| Images | Full-resolution originals with EXIF kept or stripped as agreed |
| Context shots | At least one wide view per asset visit |
| Asset record | Asset class, make and model if licensable, age band, duty cycle |
| Work-order link | Order ID, symptom, finding, action, parts replaced |
| Condition label | Supplier code plus a mapping to the buyer's scheme |
| Visit linkage | Asset ID and date so the same item can be tracked over time |
| Capture notes | Device type, lighting (flash, ambient), operator role |
| Split fields | Site, technician group and asset group for held-out evaluation |
If your use is a robot approaching equipment, add a second request for images taken at the working distance and angle of the robot's cameras, which means new capture rather than archives.
Quality and acceptance pointers
- Linkage rate. What share of images map to an asset and a work order?
- Label provenance. Is the condition code the technician's, an auditor's or inferred from text?
- Class balance. Check how many images show early-stage wear compared with failed parts; ask whether failed parts dominate the archive simply because they were the ones photographed.
- Duplicates and burst shots. Near-identical photos from one visit can leak across train and test splits.
- Resolution and focus. Sample a few hundred and inspect them, rather than relying on metadata.
- Time spread. Archives from one season may share lighting and dirt patterns.
Limitations and additional preparation
Check whether photos taken for repair justification skew toward damage that was acted on. If healthy equipment, normal wear that was left alone and borderline cases are under-represented, a model may over-call damage. Ask for matching images of equipment in acceptable condition. A related issue appears when damage was caused during handling; see handling-induced damage.
Preparation usually involves de-duplicating, joining images to records, mapping the supplier's condition codes to a scheme and removing identifying material. Nameplates, serial numbers, customer signage, license plates and faces can appear in frame; removing them can also remove useful context, and it does not remove the supplier's need to confirm it can license the images. If the equipment belongs to a customer, the supplier may need that customer's permission.
Using the images with other record types
Condition images become more valuable when combined with other maintenance data. A photo joined to a work order tells you what was found and what was done; joined to a later work order on the same asset, it tells you whether the fix held. That is the same logic used for inspection images linked to outcomes in manufacturing, applied to equipment.
For tool wear specifically, tool condition recognition covers the case where the object being judged is the tool used for the job, not the equipment being serviced. The imaging needs may differ: check whether tool wear in the archive was photographed up close on a bench and equipment condition in place, in poor light or awkward positions.
Plan the label scheme early. Supplier condition codes ("worn", "serviceable", "replace") are coarse and relative to what the technician was expected to do. If you need graded wear stages, budget for relabeling a sample with reference photos and written criteria for each stage.
Questions to ask suppliers
- Who took the photos, with what devices, and for what purpose?
- What share of images link to an asset ID, a work order and a finding?
- How were condition codes assigned, and were they audited?
- Do you hold images of equipment in acceptable condition, or only of defects?
- Are there repeat photos of the same asset across visits?
- Which images show customer names, nameplates, plates or people?
- Does your contract with the equipment owner allow licensing the images for training?
- Can you deliver a sample of a few hundred images with records for review before a license is agreed?
Discuss your requirements
If you need condition imagery tied to maintenance records, describe the equipment classes, image conditions 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. Start from the humanoid training data guide or the inspection, defect and failure data hub, and see the catalog pages on images and inspection photos and field service maintenance records. For images tied to dispositions, see inspection images linked to outcomes.
Explore sourcing and licensing options with SourceX
Sources
- MaintNet: A Collaborative Open-Source Library for Predictive Maintenance Language Resources — logbook text resources from aviation, automotive and facilities domains.