Debug File Upload Instructions for dvi3v project
v3.9.2.208

Run dvi3v with the -d or --debug flag:

dvi3v -d input.jpg output.svg

The debug output will show the temporary directory location:

Debug mode: temporary files in /tmp/dvi3v.XXXXXX

From the debug directory shown above:

1. debug_0_paper_detect.png - reference paper detection result (present only if paper detection succeeded, rather than falling back to EXIF estimation or --reference-mm)
2. debug_1_binary.png - initial binary threshold image
3. debug_2_cropped.png - cropped object region
4. With --no-holes only:
   • debug_2b_step1_clean.png - edges cleaned
   • debug_2b_step2_filled.png - holes filled
   • debug_2b_step2b_despeckled.png - stray noise pixels removed, before the margin is added
5. debug_3_final.pbm - final bitmap before vectorization
6. debug_info.txt - a guided walkthrough of the above, generated alongside them

Important: Also include the exact command line and the input image.

Rather than attaching files one by one, bundle the whole debug directory (or the relevant TestResults/ entry, if the case is also saved there) with CreateTestDataArchive — see Data Exchange Workflow for the archive tools and naming convention.

Expected: The detected reference paper region WHITE, everything else BLACK, roughly matching the paper's actual position in the photo

Purpose: Verifies which region was used to establish scale (px/mm), and by extension what the reported real-world dimensions are based on.

Common issues:

• Wrong region selected: a background area was mistaken for the paper (try --paper-color if the paper's color wasn't detected automatically)
• File absent: paper detection did not find a valid region at all (check whether an EXIF-estimated or --reference-mm scale was used instead)

Expected: Object appears WHITE, background appears BLACK

Purpose: Verifies threshold detection worked correctly.

Common issues:

• All black or all white: threshold value needs adjustment
• Multiple white regions: multiple objects detected
• Fragmented object: threshold too sensitive

Expected: Just the object region, cropped from full image

Purpose: Verifies object localization and boundary detection.

Common issues:

• Too much background included: bounding box calculation error
• Object cut off: object detection missed parts
• Wrong region entirely: the reference paper (or another large, centered region) was picked instead of the intended object

Expected: Object edges cleaned via morphological operations

Purpose: Shows the state before hole-filling.

Common issues:

• Object degraded: cleaning was too aggressive for this object's fine detail

Expected: Object WHITE with internal holes filled solid

Purpose: Verifies the flood-fill hole-filling step.

Common issues:

• Holes still visible: flood-fill did not reach them (an enclosed hole should always be filled regardless of shape; if one remains, that is a bug worth reporting with this file attached)

Expected: Same as the previous file, minus any isolated few-pixel noise specks

Purpose: Confirms stray noise was removed before the margin step would otherwise inflate it into a disproportionate, disconnected blob in the output.

Common issues:

• A small dot persists here (not just in the final output): the despeckle step did not catch it, worth reporting with this file

Expected: Final binary image ready for vectorization

Purpose: Shows what potrace will convert to SVG.

Common issues:

• Excess margin: margin calculation incorrect
• Missing parts: earlier processing failed
• A separate, disconnected shape elsewhere in the frame: usually traces back to the reference paper, or noise, having been picked up as part of the object