Start with the Vision evidence, not the appearance of the final file
Use this page when an accepted camera-only source has reached an ASCAND Vision result, but the geometry is missing, distorted, noisy, fragmented, unexpectedly closed or otherwise unsuitable.
Begin earlier if the fault is already visible in the source:
- If the object leaves the frame, the camera or object moves, the coded reference is obscured, or focus and lighting change, use Capture Setup Problems.
- If transfer, submission or processing did not reach a verified result, use General Troubleshooting.
- If a correct Vision result becomes defective only after meshing, conversion, download or import, use Mesh and Export Problems.
ASCAND Vision uses structured views and image-derived evidence to estimate object shape. Silhouettes constrain which parts of a bounded three-dimensional volume can remain occupied. This is different from measuring a projected laser line, aligning independent scans or repairing a finished mesh.
A symptom can narrow the investigation, but it does not prove a cause. Missing geometry may reflect the source, object-background separation, orientation context, evidence limits or a later representation stage. Preserve the case before deciding which explanation fits.
Preserve the source, project and result as one case
Keep the original video, the Vision project record and the failing result. Together they establish the path from observation to geometry.
Record:
- the original source filename and recording date;
- the object, capture configuration and relevant setup photographs;
- the project or job identifier;
- the selected scan method and physical configuration;
- the verified processing state;
- the exact result or representation being inspected;
- the symptom and affected region; and
- every recapture, resubmission, conversion or repair already attempted.
Do not overwrite the source, replace the only result or begin by smoothing the failing geometry. Those actions can remove evidence without explaining the fault.
Save screenshots from consistent viewpoints when comparing results. A different camera angle, shading mode or representation can make the same geometry look better or worse.
If you create another attempt, give it a distinct identity. Change one relevant condition, keep the remaining workflow stable and preserve both outcomes.
Review the general evidence checklist.
Verify the accepted source and Vision project lineage
Before diagnosing reconstruction, confirm that the source still passes the complete acquisition check.
Verify:
- the complete object and coded turntable remain framed throughout rotation;
- the camera remains fixed;
- the object remains rigid and stationary relative to the turntable;
- focus, exposure and illumination remain usable;
- object and background remain visually separable;
- the coded reference remains sufficiently visible under the applicable instructions;
- the recording is continuous and belongs to this object; and
- the unedited original is available.
Then verify the project:
- The submitted file is the accepted original.
- Vision was the intended scan method.
- The selected physical configuration matches the capture.
- Transfer reached its verified completed state.
- The Vision job was accepted and reached a documented result state.
- The inspected result belongs to that job.
Upload completion, job acceptance, processing completion and result suitability are separate gates. A completed job proves neither correct source interpretation nor suitable geometry.
If the project lineage is uncertain, do not diagnose shape from the result alone. Return to the Vision workflow and current platform documentation.
Check orientation context and silhouette quality
ASCAND’s controlled rotation supplies a sequence of observations around the object. The coded turntable helps associate those observations with rotational context. Vision processing can then use the object’s separation from the background to constrain possible occupied volume.
Ask two groups of questions.
Is the orientation evidence consistent?
- Did the coded reference remain visible and clear across the rotation?
- Did the object remain fixed relative to the turntable?
- Was the correct physical configuration selected?
- Is there an interruption, reversal or unexpected change in the recording?
Is the object boundary dependable?
- Does the object remain visually distinct from the background?
- Do shadows merge with the object boundary?
- Do reflections or transparent regions change appearance with orientation?
- Do thin features disappear against the background?
- Does a support become part of the visible outline?
A clean outline in one frame is not enough. Vision geometry depends on mutually consistent evidence across orientations.
Do not assume access to internal silhouettes, masks, thresholds or confidence data. If the current interface exposes verified diagnostic evidence, interpret it using current documentation. Otherwise, inspect the observable source conditions rather than inventing an internal failure.
Learn how silhouettes constrain occupied volume.
Diagnose missing, swollen, flattened or distorted regions
Use the visible shape to formulate an evidence question.
| Symptom | Evidence question |
|---|---|
| A real feature is missing | Was it consistently visible and separated from the background, or could valid volume have been contradicted by the observations? |
| Extra or swollen volume remains | Did background, shadow, support or uncertain separation appear as part of the object region? |
| A concavity is filled | Did that recess ever change the visible outline, or is it outside what silhouette evidence can determine? |
| The underside is incomplete | Was that region observed in this orientation, or does it require a complementary capture? |
| A thin part is shortened or lost | Did it remain stable and visually distinct in enough views? |
| The whole form is skewed or flattened | Are source identity, orientation context, object stability and configuration all verified? |
These are diagnostic questions, not automatic cause assignments.
Silhouette-based reconstruction estimates a visual hull. A concavity that never affects any observed outline may remain filled because the source does not contradict that volume. A closed surface can therefore look coherent while still simplifying or misrepresenting the object.
Likewise, a gap does not always mean that a mesh should be patched. If the missing region was never adequately observed, repair creates new geometry; it does not recover evidence.
Review Vision strengths and limits.
Separate noise, fragments and rough boundaries from later mesh defects
Identify which representation first contains the symptom.
- Occupied-volume or voxel-derived result: investigate source consistency, boundary evidence, coverage and reconstruction limits.
- Extracted surface or mesh: determine whether the underlying volume is sound before diagnosing topology, normals, holes or surface extraction.
- Converted or downloaded file: compare it with the platform result to separate conversion, transfer and destination-tool interpretation.
Detached fragments may represent background, support or shadow retained as possible occupancy. Rough boundaries may reflect uncertain silhouette separation or the spatial representation used to express the occupied boundary. Thin structures can appear discontinuous when the available evidence does not constrain them consistently.
Do not assume that smoothing, decimation, remeshing, hole filling or changing file format will correct the evidence. These operations can alter representation and topology. They cannot prove where the original surface belonged.
Also avoid reading a block-like or stepped boundary as proof of one specific internal resolution. The current public workflow does not establish a user-visible voxel setting merely from result appearance.
If the source and Vision result are sound but the mesh, export or imported file is not, continue with Mesh and Export Problems.
Change one evidence condition and compare the same region
Use a controlled comparison only after preserving the baseline.
- Select one observable symptom and one affected region.
- Confirm that the original source and result are identifiable.
- Choose one relevant source-level condition supported by current instructions.
- Record a new complete source while holding other conditions stable.
- Submit it as a distinct Vision project using the verified workflow.
- Compare the same representation, viewpoint and object region.
- Record what changed, what remained and whether the symptom moved.
Appropriate changes might involve correcting a confirmed framing, stability, boundary-separation or visibility problem. They do not include guessed internal thresholds or undocumented processing controls.
Repeatedly processing the same unchanged source may test repeatability, but it does not add missing observations. Changing background, lighting, object placement and camera framing together may produce a different result, but it cannot identify which condition mattered.
Treat improvement as evidence supporting a hypothesis, not universal proof. Preserve unsuccessful tests too; they narrow the next decision.
Use the controlled-comparison method.
Recapture, choose complementary evidence or escalate
Choose the next owner from the earliest verified limit:
- A source defect is visible: correct one acquisition condition and recapture through Capture Setup Problems.
- The source is sound, but the object exceeds what its silhouettes establish: review object suitability and compare Vision, Laser and Combo.
- The symptom belongs to laser-enabled evidence or Combo reconciliation: use Laser and Combo Scan Problems and follow current safety instructions.
- The Vision geometry is acceptable, but a mesh or export is not: use Mesh and Export Problems.
- The case remains unexplained: contact Support with the preserved evidence set.
For escalation, provide the source and project identities, applicable configuration, exact result, affected region, screenshots from repeatable viewpoints, processing state and a list of controlled tests. Share project files only through the approved support route and according to the governing privacy guidance.
Do not claim that a different method will automatically improve every region. Vision, Laser, Combo and complementary orientations provide different evidence. The correct next path depends on the object, the unresolved geometry and the intended use.