Vehicle Damage Assessment

Analyze vehicle damage photos with AI-powered vision analysis. Detect dents, scratches, cracks, and estimate repair complexity — perfect for insurance claims and auto body shop assessments.

Features

• Multi-part analysis: Identify damage location, type, and severity
• Repair complexity estimation: Low, medium, or high complexity assessment
• Safety concerns detection: Identify structural, airbag, and glass hazards
• Confidence scoring: AI confidence level (0.0 to 1.0) for each analysis
• Per-damage breakdown: Every detected damage becomes its own row, with a derived repair action and urgency priority
• Batch processing: Analyze up to 20 vehicle photos in one run
• Local AI processing: Qwen2.5-VL vision model via ntriq AI infrastructure

Input

Pass an array of image URLs — one row per photo plus one row per damage found:

{
  "imageUrls": [
    "https://example.com/vehicle-1.jpg",
    "https://example.com/vehicle-2.jpg"
  ],
  "maxImages": 20
}

A single photo also works:

{
  "imageUrl": "https://example.com/vehicle-damage.jpg"
}

Automation clients do not have to match those field names exactly. image_url, image_urls, images, photos, url, urls and startUrls are all read, each item may be a plain string or an object with a url field, and a missing https:// is filled in. Values that are not image URLs are skipped and reported in a notice row instead of ending the run.

Duplicate URLs are removed before analysis, so you are never charged twice for the same photo. If you supply more URLs than maxImages, the extras are skipped and reported in a notice row instead of failing the run.

Output

Each photo produces one assessment row summarizing the whole vehicle, one damage_item row per detected damage, and one panel_inspection row for every other part the model could see — so an undamaged photo still comes back as a per-part condition checklist rather than a single "no damage" line. A part reported as damaged appears once, as a damage_item row.

The Actor downloads each photo itself with a declared bot User-Agent and passes the bytes to the vision model. Hosts that reject anonymous server-side fetches (Wikimedia and most CDNs answer those with 403) therefore work; the row's imageSource field records which path was used.

Assessment row:

{
  "rowType": "assessment",
  "imageUrl": "https://example.com/vehicle-damage.jpg",
  "imageSource": "actor_fetch",
  "status": "success",
  "parseStatus": "ok",
  "vehicleDetected": true,
  "vehicleView": "front",
  "assessmentNote": "",
  "damageDetected": true,
  "overallSeverity": "moderate",
  "severityScore": 2,
  "damages": [{ "area": "front_bumper", "damageType": "dent", "severity": "moderate", "description": "Significant dent in center of front bumper" }],
  "damageCount": 2,
  "damageAreas": ["front_bumper", "hood"],
  "damageTypes": ["dent", "scratch"],
  "structuralDamage": false,
  "highestPriority": 2,
  "panelsInspected": 6,
  "panelsIntact": 4,
  "estimatedRepairComplexity": "medium",
  "safetyConcerns": false,
  "safetyNotes": "",
  "confidence": 0.87,
  "model": "Qwen2.5-VL",
  "processingTimeMs": 2341
}

Damage row:

{
  "rowType": "damage_item",
  "imageUrl": "https://example.com/vehicle-damage.jpg",
  "status": "success",
  "damageIndex": 1,
  "area": "front_bumper",
  "damageType": "dent",
  "severity": "moderate",
  "severityScore": 2,
  "description": "Significant dent in center of front bumper",
  "repairActionHint": "paintless dent repair or panel beating",
  "structuralArea": false,
  "repairPriority": 2
}

Part inspection row:

{
  "rowType": "panel_inspection",
  "imageUrl": "https://example.com/vehicle-damage.jpg",
  "status": "success",
  "panel": "rear_bumper",
  "condition": "intact",
  "structuralArea": false,
  "notes": "No visible damage on the rear bumper"
}

Output Fields

Field	Type	Description
rowType	string	assessment, damage_item, panel_inspection, or notice
imageUrl	string	Original image URL
imageSource	string	actor_fetch when the Actor downloaded the photo, vision_service when the URL was passed through
vehicleDetected	boolean	Whether a vehicle was visible at all (null when the reply could not be parsed)
vehicleView	string	front, rear, side_left, side_right, interior, top, or unknown
assessmentNote	string	Why the photo could not be assessed — set when no vehicle was visible
panel / condition / notes	string	Per-part detail (panel_inspection rows): part name, intact/damaged/unclear, and a short observation
panelsInspected / panelsIntact	number	Distinct visible parts, and how many of them are intact (assessment rows)
status	string	"success" or "error"
parseStatus	string	ok, or raw_text_fallback when the model replied in free text (see rawAnalysis)
damageDetected	boolean	Whether damage was found (null when the reply could not be parsed)
overallSeverity	string	"none", "minor", "moderate", "severe", or "total_loss"
severityScore	number	Numeric severity for sorting: none 0 → total_loss 4
damages	array	Full damage array as returned by the model
damageCount	number	Number of distinct damages in the photo
damageAreas / damageTypes	array	Distinct parts and damage types found
structuralDamage	boolean	True when a structural or safety-critical part is affected
highestPriority	number	Most urgent priority in the photo: 1 act now, 3 cosmetic
area / damageType / severity	string	Per-damage detail (damage_item rows)
repairActionHint	string	Suggested repair approach derived from the damage type
repairPriority	number	1 act now, 2 schedule soon, 3 cosmetic
estimatedRepairComplexity	string	"low", "medium", or "high"
safetyConcerns	boolean	Safety hazards identified
safetyNotes	string	Details of safety issues (structural, airbag, glass, steering)
confidence	number	AI confidence 0.0-1.0
model	string	Vision model identifier
processingTimeMs	number	Processing time in milliseconds
code / httpStatus	string / number	Failure detail when status is "error"

When a photo cannot be analyzed — the image host blocks the download, or the vision service times out — you get an error row carrying code and httpStatus so you can see exactly what happened. Failed photos and notice rows are not charged.

Pricing

Free plan: each run returns up to 25 result rows (the first 3 photos). Paid Apify plans receive the full result set.

$0.20 per analyzed photo — charged once per photo regardless of how many damage rows it produces. Photos that could not be analyzed are not charged.

Supported Damage Types

• dent — Impact deformation
• scratch — Surface abrasion
• crack — Fracture in material
• shatter — Broken glass
• deformation — Bent or warped
• paint_damage — Paint loss or chipping
• rust — Corrosion

Supported Vehicle Parts

Common areas analyzed include:

• Front/rear bumpers
• Hood
• Windshield/windows
• Doors (front/rear, left/right)
• Fenders (front/rear)
• Quarter panels
• Side mirrors
• Taillights/headlights
• Trunk/tailgate
• Roof

Legal Notice

NOT a Professional or Insurance Assessment (IMPORTANT)

• This tool provides AI-assisted PRELIMINARY damage analysis for INFORMATIONAL PURPOSES ONLY.
• Results are NOT a substitute for professional vehicle inspection or certified damage appraisal.
• DO NOT use this tool's output for insurance claims, legal proceedings, or official valuations.
• Insurance claims require assessment by a licensed insurance adjuster in your jurisdiction.
• Using AI-generated damage estimates in insurance claims may result in claim denial or fraud allegations.

Before You Upload

• REMOVE or BLUR license plates, VIN numbers, and any personally identifiable information from photos before uploading.
• Vehicle registration numbers and VINs are considered personal data under GDPR and CCPA.
• We do not extract, store, or process license plate or VIN information, but you should redact them as a precaution.

Accuracy Limitations

• AI analysis accuracy varies significantly by image quality, angle, lighting, and damage type.
• The model cannot detect hidden, internal, or structural damage not visible in photos.
• Severity estimates and repair complexity ratings are rough approximations only.
• Always consult a certified mechanic or body shop for actual repair assessments.

Liability

• The developer assumes NO liability for any decisions made based on this tool's output.
• This includes but is not limited to: insurance claims, repair decisions, vehicle purchases, safety determinations, and legal proceedings.
• Users are solely responsible for verifying results with qualified automotive professionals.

Data Processing

• Vehicle images are processed on our local AI server and immediately discarded after analysis.
• No images are stored, cached, or shared with third parties.
• We do not retain any vehicle identification information.

Use Cases

✅ Recommended:

• Initial damage assessment for insurance claims
• Quick reference for auto body shops
• Vehicle condition documentation
• Fleet management damage tracking
• Quick quotes before professional inspection

❌ Not Recommended:

• Official insurance appraisals (use certified adjusters)
• Safety certifications or compliance (use inspectors)
• Legal/liability determinations (use professionals)
• Single image reliance (multiple angles recommended)

Example Workflow

1. Photograph vehicle damage from multiple angles
2. Provide URLs to clear, well-lit images
3. Receive AI assessment with confidence scores
4. Verify results with a professional inspector
5. Use for preliminary documentation or quick estimates

Support

For questions or issues, contact: support@ntriq.co.kr

---

Remember: This is AI-assisted analysis for informational purposes. Always consult qualified professionals for important decisions.

---

🔗 Related Actors by ntriqpro

Extend this actor with the ntriqpro intelligence network:

• video-intelligence-analyzer — Video/image AI analyzer (transcripts, scenes, objects)
• table-chart-extractor — Table and chart data extraction
• image-batch-optimizer — Bulk image optimization (resize/compress)

⭐ Love it? Leave a Review

Your rating helps professionals discover this actor. Rate it here.