Automate damage assessment
CV engineers train damage-detection and severity models on 1B+ labeled photos straight from real auction lanes.
Skip months of in-house labeling
— fine-tune off real auction declarations.
Image & Vision Data
Vehicle image and damage datasets for computer vision
Over 1 billion labeled photos of vehicle damage, plus 1.5 billion listing images — every photo tied to a VIN, year, make, and model. Built to train damage detection, condition grading, and parts-recognition models.
1B+
labeled damage images
1.5B+
listing images
2.5B+
images total
VIN-level
labels
+8M
images per day
What's in the data
Labeled fields on every image
Photos arrive with the labels a vision model actually needs — not a raw folder of pictures. Both new and used vehicles, every brand, every region.
Damage location
— which panel the damage is on.
Damage type
— dent, scratch, crack, glass, or structural.
Severity grade
— how bad the damage is, calibrated to auction grades.
Photo source
— auction lane, dealer lot, or listing.
YMMT label
— year, make, model, and trim.
VIN link
— every photo joins back to the vehicle's record.
{
"vin": "1FTFW1E84MFA12345",
"image_url": "https://img.vinassessment.com/dmg/1FTFW1E84MFA12345/07.jpg",
"damage": [
{ "panel": "front_bumper", "type": "crack", "severity": "moderate" },
{ "panel": "left_fender", "type": "dent", "severity": "minor" }
],
"source": "auction_lane",
"ymmt_label": "2022 Ford F-150 Lariat SuperCrew 4x4",
"captured_at": "2026-04-18T14:22:08Z"
}
Who Uses It
Who builds on these images, and why
Real vision pipelines, trained on labeled photos at a scale that's hard to assemble in-house.
Estimate repairs from a photo
Estimation tools learn to map visible damage to repair operations and cost from labeled panel, type, and severity.
Quote a repair from a single image
.
Score and flag listing photos
Marketplaces auto-score photo quality and flag visible damage that buyers should see up front.
Cleaner listings, fewer disputes
.
Automate condition grading
Platforms train models that grade condition from lane photos, calibrated to real auction condition grades.
Consistent grades at scale
.
Delivery
How you get the images
Images and their labels arrive in your infrastructure, ready to load into a training pipeline.
Image archive
Signed S3 URLs to the photo set — pull the images straight into your training jobs.
Metadata in Parquet / JSON
Labels and VIN links alongside the images, so every photo joins back to its vehicle.
Filtered subsets
By make, damage type, or severity — pay for what you need.
What Makes It Different
Labeled, linked, and at real scale
Three things you'll notice the first time you compare this to a generic image scrape.
Labeled, not raw
Every image carries panel, type, and severity — not a folder of photos you still have to annotate.
Tied to real VINs and titles
Each photo links to a VIN, so you can cross-join to listings, auctions, and title brands.
Damage at a scale few others have
1B+ damage photos from real auction lanes — breadth that's hard to assemble from scratch.
How You Buy It
Buy only what you need — clean and ready to use
No all-or-nothing contract and no raw scrape to clean up. Take the exact slice you need, already normalized and ready for your pipeline.
Clean data, not a raw scrape
Every record is normalized, VIN-keyed, deduplicated across sources, and validated — with a documented, versioned schema. It drops straight into your warehouse or training pipeline, with no cleanup pass on your side.
Buy it in pieces
Take the whole dataset or just the slice you need. License one dataset or several combined, as a one-time pull or an ongoing feed — and always start with a sample.
make & model
region / state
year range
date range
just the fields you need
FAQ
Frequently asked questions
Each damage image carries the panel (where on the car), the damage type, and a severity grade. Labels are sourced from real auction declarations and condition reports, then reviewed for accuracy before delivery.
Yes. The license covers commercial computer-vision and ML model training. Some image sources carry attribution requirements your account team will flag during scoping.
You get the image archive via signed download URLs plus the labels and VIN links as metadata in Parquet or JSON, pushed to your S3 bucket or SFTP endpoint. No portal to log into.
Yes — filter by make, damage type, or severity and pay only for the slice you need. We deliver a sample set first so you can validate the labels before committing.
Get Started
Tell us what you need
Tell us about your computer-vision use case and we'll come back with a dataset recommendation, a sample, and a quote — usually within 2 business days. No sales call required to get the numbers.
A billion labeled photos, ready to train on
1B+ labeled damage images · 1.5B listing images · VIN-level labels · delivered to your cloud. New and used, every brand, every region. Tell us what you're building.