Photo set checker
Photogrammetry usually fails because of the photos, not the software — and you find out an hour into reconstruction. Drop your photo set here first and find out in seconds which frames are blurred, blown out, or too dark to match, and whether you shot enough of them. Runs entirely in your browser.
About this capture checker
Photogrammetry turns a set of ordinary photos into a 3D model by finding the same physical features across many shots and solving where the camera must have been. That process is unforgiving about input: a blurred frame has no crisp features to match, a blown-out one has no detail at all, and a set with gaps has nothing to connect one side of the object to the other.
The expensive part is that you normally find this out after reconstruction, which can take an hour or more on a good GPU. This checks the same things in seconds and names the specific photos to delete or reshoot.
Your photos are never uploaded. Each one is measured inside your browser. Sharpness is judged relative to the sharpest frame in your own set rather than against a fixed threshold, because what counts as sharp depends entirely on your camera, your subject, and your light.
Common questions
- How many photos do I need for a 3D scan?
- At least 20 overlapping shots for the solve to work at all, and 40 or more before results get reliable. Each photo should overlap the previous one by roughly two-thirds — the software connects them through shared features, so gaps break the chain.
- Why does my photogrammetry keep failing?
- Most often motion blur from handheld shooting, not enough overlap between shots, or a featureless subject. Shiny, transparent, and plain single-colour surfaces give the matcher nothing to lock onto, no matter how good the photos are.
- How does it decide a photo is blurry?
- It measures the variance of the Laplacian, which is the standard sharpness metric. A blurred image has little fine detail, so its second derivative is small everywhere. Frames are then compared against the sharpest photo in your own set.
- Should I delete the bad frames or just leave them?
- Delete them. A bad frame is worse than a missing one — it can be matched incorrectly and pull the whole camera solve off, which degrades the model everywhere, not just at that angle.
- What camera settings work best?
- Small aperture for depth of field, fast enough shutter to avoid motion blur, fixed focus and exposure across the whole set, and diffuse even lighting. Consistency between frames matters more than any single frame being beautiful.
- Can it build the 3D model too?
- Not in a browser. Dense reconstruction needs sustained GPU compute well beyond what a tab can do. This tool makes sure the input is worth reconstructing; the reconstruction itself is a desktop job.