Help Center → Capture

Understanding Photo Quality Requirements

The AI that analyzes your inspection photos has specific requirements for lighting, sharpness, angle, and coverage. Here's what it looks for and what gets flagged.

6 min read


Understanding Photo Quality Requirements

Estimatics runs every photo through a quality check before AI analysis. Photos that don't meet the minimum requirements are flagged — they won't produce accurate damage findings, and in some cases they won't be processed at all. Understanding these requirements helps you capture better evidence from the start.


How photo quality is evaluated

When a photo uploads, Estimatics automatically checks for:

  1. Sharpness — Is the subject in focus?
  2. Exposure — Is the photo properly lit (not too dark or overexposed)?
  3. Resolution — Is the image large enough for AI analysis?
  4. Angle — Is the camera angle usable for the subject type?
  5. Coverage — Does the photo show the claimed subject clearly?

Photos that pass all checks move directly to AI analysis. Photos that fail one or more checks are flagged with a specific reason so you can retake or supplement them.


Quality flags and what they mean

FlagCauseFix
Blurry / Out of focusCamera motion or subject too closeTap to focus before shooting; hold steady
UnderexposedToo dark — detail lost in shadowsTap the bright area to adjust exposure; use flashlight for interiors
OverexposedToo bright — detail lost in highlightsTap a bright surface to expose correctly; avoid shooting into direct sun
Low resolutionFile too small for analysisAvoid screenshots; use the native Estimatics camera
Extreme angleMore than ~60° from perpendicularShoot more directly at the surface
Subject unclearDamage area not identifiableRetake closer or at a different angle
DuplicateNear-identical to another photo in the same areaRemove one; duplicates don't add evidence value

Flagged photos are still stored in the job — they're just marked for your review. You can override a flag if you believe the photo is usable.


Minimum technical requirements

RequirementMinimumRecommended
Resolution4MP8MP+ (iPhone default is 12MP+)
FocusSubject in focusTap-to-focus on damage surface
ExposureKey damage detail visibleProperly exposed for the surface, not the sky
Angle from perpendicularLess than 60°Less than 45° for measurement accuracy
Subject in frameDamage area occupies >15% of frame25–50% for close-up detail shots

The three-distance rule

The most common quality issue isn't technical — it's coverage. Inspectors often submit only close-up shots without the context shots that make them meaningful.

For every significant damage area, capture photos at three distances:

Establishing shot (15–30 feet away) Shows the full surface — the entire roof plane, the whole elevation, the complete room. Gives context for where the damage exists within the property.

Mid-range shot (3–10 feet away) Shows the damage area within the surface. The viewer can see both the damage and enough surrounding area to understand its extent and location.

Close-up shot (1–3 feet away) Shows the specific damage detail the AI needs to classify — impact marks, granule displacement, crack patterns, water staining boundaries, mold growth, etc.

All three distances together are significantly more defensible than close-ups alone. A carrier can't dispute damage they can see at all three levels.


Common photo problems in the field

Shooting directly into sunlight

Problem: Backlighting makes damage invisible even when it's clearly present. Fix: Reposition to put the sun at your back, or wait for the sun to move. For roof shots from a ladder, shoot in the early morning or late afternoon.

Dark interiors

Problem: Room lights are off and natural light is insufficient — damage details lost in shadows. Fix: Turn on every light in the room. Supplement with your iPhone's flashlight (swipe up to Control Center, enable flashlight, then open the camera). Tap the darkest area of the frame to expose for it.

Ladder shake

Problem: Photos taken from a ladder while holding the device one-handed are frequently blurry. Fix: Brace your elbow against your body or the ladder. Use the volume button to trigger the shutter instead of tapping the screen — it produces less camera movement.

Wet or dirty lens

Problem: Fingerprints and moisture on the camera lens reduce sharpness across the entire photo. Fix: Wipe the lens with a clean cloth before starting any inspection session. Check again after handling gutters or dirty surfaces.

Screenshots of paper documents

Problem: Some inspectors photograph paper sketches, printed reports, or handwritten notes. Fix: Don't include these in the Capture module — they don't add AI-analyzable evidence. Use the Notes section for text context.


Reviewing flagged photos

After uploading, go to the Capture tab on the web dashboard:

  1. Click the Flagged filter to see only photos with quality issues
  2. Review each flagged photo and the reason
  3. If the photo is still useful, click Keep Anyway to override the flag
  4. If it needs to be replaced, go back to the field (or the best available alternative) and re-shoot

Best practice: Do your quality review before leaving the property. It takes 5 minutes and saves a return trip.


Frequently asked questions

Will flagged photos affect my AI findings? Flagged photos that are kept in the job will still be analyzed, but with lower confidence. The AI notes which findings came from lower-quality photos. For CERTIFY plans, flagged photos reduce the Defendibility Score.

Can I manually mark a photo as high-priority for the AI? Not directly. The AI processes all photos in an area. To improve findings for a specific area, add more high-quality photos to that area.

My photos look fine on my phone but get flagged — why? iPhone screens are very high resolution and bright, which makes most photos look good on-device. The AI is analyzing the actual pixel data, which can reveal sharpness or exposure issues not obvious on a phone screen. Compare on a computer monitor for a more accurate assessment.

How many flagged photos are acceptable? For launch: zero. Aim to have no flagged photos by the time you close a job. For CERTIFY plans, a high number of flagged photos will visibly lower the Defendibility Score.



Last updated: March 2025 · Questions? Use the Resources panel in the app or email support@aiestimatics.com

Last updated: March 2025 · Feedback on this article