Photo Best Practices

Good photos lead to good AI detections, and bad photos lead to missed defects and false positives. This article covers the practical photo capture techniques that will get you the best results from SuperClear AI. These aren’t strict rules — they’re proven habits that experienced inspectors have developed for working with computer vision.

The Golden Rules

Sharp focus — Tap to focus before you take the shot
Good lighting — Avoid heavy shadows and backlighting
Get close enough — The WRB should fill most of the frame
Straight-on angles — Avoid extreme tilts when possible
One condition per photo — Capture each defect area separately

Stick to these five and you’ll dramatically improve your detection accuracy.

Focus

Computer vision models depend on sharp edges and clear textures. A blurry photo can hide defects entirely or trigger the wrong predictions.

Tap to focus on your phone’s camera before tapping the shutter — most cameras let you tap the area of interest to lock focus
Hold still for a moment after tapping the shutter — moving as you click is the most common cause of blur
Wait for the autofocus indicator (the box or square) to confirm focus
If a photo is blurry, retake it — don’t try to compensate with editing later

Lighting

Natural light is best

Daylight gives the most consistent, even lighting for inspection photos. Early morning and late afternoon avoid the harsh shadows of midday sun. Overcast days are surprisingly good — soft, even light without strong shadows.

When to use flash

Indoors with poor lighting
Behind walls or in shadows where natural light doesn’t reach
Early morning before sunrise or late evening

Be aware that flash can wash out subtle defects and create harsh reflections on shiny surfaces. If you have a choice, natural light wins.

What to avoid

Backlighting — Don’t shoot toward a bright light source. The camera will expose for the bright background and your subject will be too dark.
Direct overhead sun — Casts hard shadows on textured surfaces
Mixed lighting — Half of the frame in shadow and half in sunlight makes the AI’s job harder

Framing and Distance

Fill the frame

The WRB area you’re inspecting should be the main subject of the photo, filling most of the frame. If the area of interest is just a tiny part of a wide shot, the AI may not recognize details. Get closer.

How close is too close?

Too close — The camera can’t focus, or you can only see a small section of fabric without context
Just right — The defect is clearly visible and you can see the surrounding context (a few inches around the defect)
Too far — The whole wall fills the frame and the defect is one tiny part

When in doubt, take two photos — one closer and one with more context. You can always delete the one you don’t need.

Angle

Capture the WRB straight-on when you can. Heavy angles distort the appearance of seams, fasteners, and tape, and can make a properly installed area look defective. If you can’t get straight-on, try a shallow angle (within ~30 degrees of perpendicular).

One Condition Per Photo

Take a separate photo for each defect or area of concern, instead of trying to capture multiple issues in a single wide shot.

Why this matters:

The AI can mark each photo with its specific findings
Photo tags work better when each photo represents one condition
Punch lists are clearer when each finding has its own image
You can delete or fix individual photos without losing context

Removing Obstructions

Move scaffolding, ladders, tools, and hands out of the frame when possible
Wait for workers to step out of the shot
If something is unavoidable, take a second photo from a different angle that isn’t obstructed

Capturing Context

When you find an unusual condition, take both a close-up and a wider context shot:

Close-up — Shows the defect clearly for AI detection
Wide shot — Shows where the defect is in the building (helpful for the report reader)

Tag both photos with the same condition so they group together.

Add Metadata While It’s Fresh

After uploading, take a few seconds to add building, floor, elevation, and a caption to each photo. The AI doesn’t need this metadata to detect defects, but it makes the punch list immediately useful and lets you filter photos later.

Building — Pick from the project’s building options
Floor — Pick from the project’s floor options
Elevation — North, South, East, or West
Caption — A short description of what you’re seeing

See Managing Photo Tags and Conditions for more on tagging.

Batching Captures

When walking a site, capture all the photos for one area before moving on. This gives you cleaner mental groupings later when you review and tag everything.

When to Re-shoot

If you upload a photo and the AI seems to have missed something or flagged something incorrectly:

First, check the photo quality — was it sharp? Well-lit? Properly framed?
If yes, the AI may have legitimately missed a defect. Use the image editor to add your own annotation.
If the photo quality was poor, retake it with the techniques above and re-upload

Bad Habits to Break

Snapping without focus — Always tap to focus first
Zooming in digitally — Move closer instead, digital zoom just crops a smaller area at the same resolution
Wide panoramic shots of entire walls — Useful for context, useless for AI detection
Heavy editing before upload — Filters and edits can confuse the AI. Capture the photo as it actually is.
Mixing multiple defects in one shot — Take separate photos for each finding

Device Tips

iPhone: Use the standard Camera app settings — “High Efficiency” format is fine
Android: Most modern Android cameras work well; avoid extreme zoom
iPad: Excellent for inspection photos — large viewfinder makes framing easier
Battery and storage: Bring a power bank and check storage before going to a long site visit