Testing Plan for dvi3v project
v3.9.2.208

Establish system works correctly under ideal conditions.

1. Simple Geometric Shapes

   • Black circle (25mm, 50mm, 75mm diameter)
   • Black square (25mm, 50mm, 75mm sides)
   • Black rectangle (50mm × 100mm)
   • Black triangle (equilateral, 80mm sides)

   Why: Known dimensions for accuracy validation Expected: ±1mm accuracy, clean contours

2. Reference Standards

   • Printed reference card with known dimensions
   • Calibration target (checkerboard pattern)

   Why: Establish baseline measurement accuracy Expected: Exact dimensional match

• Measure physical objects with calipers (±0.1mm)
• Compare SVG dimensions with physical measurements
• Calculate error percentage
• Document scale detection method (A4 vs base-feet when implemented)

• Dimensional error < 2% for all baseline objects
• Contour completeness 100% (no gaps)
• Consistent scale detection across multiple captures

Test detection across different materials and contrast levels.

1. Matte Materials

   • Matte black paper cutout
   • Matte colored paper (red, blue, green)
   • Cardboard (brown)

   Why: No reflections, ideal for detection Expected: Near-perfect extraction

2. Glossy Materials

   • Glossy black paper
   • Laminated card
   • Photo print

   Why: Reflections may confuse edge detection Expected: May need lighting adjustment or threshold tuning

3. Textured Materials

   • Fabric swatch
   • Textured cardboard
   • Cork

   Why: Texture may create noise in binary threshold Expected: May need morphological operation tuning

4. Transparent/Translucent

   • Clear plastic shape
   • Frosted plastic
   • Colored acetate

   Why: Edge case - minimal contrast Expected: Likely to fail, documents limitations

5. Metallic/Reflective

   • Aluminum cutout
   • Copper shape
   • Chrome-plated object

   Why: Specular reflections, extreme edge case Expected: High failure rate, documents limitations

✓ = Required test, - = Optional

• Document threshold value needed for successful extraction
• Note any preprocessing required (lighting adjustment)
• Measure extraction quality score
• Photograph failure cases for documentation

Validate handling of increasingly complex shapes.

1. Level 1 - Convex Shapes (No Holes)

   • Circle, square, triangle (already in baseline)
   • Ellipse
   • Pentagon, hexagon, octagon
   • Smooth organic shape (leaf outline)

   Expected: Perfect extraction

2. Level 2 - Concave Shapes

   • Star shape (5-pointed)
   • Crescent moon
   • Hand silhouette
   • Letter shapes (C, G, S)

   Expected: Good extraction, may need contour simplification

3. Level 3 - Shapes with Holes

   • Ring (donut shape)
   • Letter O, B, P, A
   • Picture frame shape
   • Swiss cheese pattern (multiple holes)

   Expected: Requires --no-holes flag testing Critical: Known issue to validate fix for

4. Level 4 - Thin Features

   • Scissors (thin blades)
   • Key (thin shaft)
   • Fork/spoon (tines/handle)
   • Wire frame object

   Expected: Thin features may disappear with threshold Test: Threshold sensitivity

5. Level 5 - High Detail

   • Gear (teeth)
   • Snowflake pattern
   • Lace doily
   • Circuit board outline

   Expected: Detail loss at coarse threshold Test: Auto-threshold feature

• Perimeter-to-area ratio
• Number of concave vertices
• Number of holes
• Minimum feature width
• Contour point count in SVG

• Visual inspection of SVG overlay on original photo
• Measure detail preservation percentage
• Test different threshold values (manual and auto)
• Document optimal threshold range per complexity level

Establish minimum and maximum object size limits.

1. Minimum Size Tests

   • 10mm × 10mm square
   • 15mm × 15mm square
   • 20mm × 20mm square
   • 5mm diameter circle
   • 10mm diameter circle

   Why: Find lower detection limit Expected: Below certain size, lost in noise or margin

2. Maximum Size Tests

   • 180mm × 250mm rectangle (nearly full A4)
   • Object extending to paper edges
   • Object larger than paper (partially off-frame)

   Why: Test margin handling and cropping Expected: Margin settings become critical

3. Aspect Ratio Extremes

   • 200mm × 10mm strip (20:1 ratio)
   • 10mm × 200mm strip (1:20 ratio)
   • Very elongated ellipse

   Why: Test bounding box calculation Expected: Should handle well but may challenge auto-cropping

• Binary search for minimum detectable size
• Test with different margin settings (0mm, 2mm, 5mm)
• Document size limits in documentation
• Verify no clipping at maximum sizes

Validate center-priority detection algorithm.

1. Center Position (Baseline)

   • Object exactly centered on A4

   Expected: Always selected correctly

2. Off-Center Positions

   • 25mm from center (4 directions: N, S, E, W)
   • 50mm from center
   • 75mm from center
   • Near edge of A4 (10mm margin)

   Expected: Should still detect if within reasonable distance

3. Multiple Objects Test

   • Two identical objects: one centered, one at edge
   • Three objects: small centered, large at edges
   • Multiple small objects scattered

   Expected: Center object selected Validation: Confirms center-priority algorithm

4. Rotated Placement

   • Object rotated 45°, 90°, 180°

   Why: Ensure rotation doesn't affect detection Expected: No impact on detection

• Photograph grid positions (3×3 grid on A4)
• Test each grid position
• Document detection success rate vs. distance from center
• Create heat map of detection reliability

Establish lighting requirements and robustness.

1. Uniform Lighting (Baseline)

   • Diffuse overhead light
   • Shadowless, even illumination

   Expected: Ideal conditions, 100% success

2. Directional Lighting

   • Single light source from left
   • Single light source from above
   • Creates shadow on one side

   Expected: Shadow may be detected as part of object Test: Shadow impact on dimensions

3. High Contrast Lighting

   • Direct sunlight
   • Very bright studio lights

   Expected: May cause overexposure or harsh shadows Test: Threshold adjustment needed

4. Low Light

   • Dim indoor lighting
   • Evening/twilight conditions

   Expected: Reduced contrast, may need exposure adjustment Test: Camera exposure compensation needed

5. Mixed Color Temperature

   • Daylight + incandescent
   • Fluorescent + LED

   Expected: Color cast may affect colored object detection Test: Impact on threshold for colored objects

• Use light meter if available
• Document lux levels
• Photograph test setup showing lighting
• Record camera exposure settings (ISO, shutter, aperture if DSLR)

• Test same object under different lighting
• Compare threshold values required
• Measure dimensional accuracy variance
• Document acceptable lighting ranges

Validate scale detection reliability.

1. Optimal A4 Placement

   • A4 fully visible, flat, parallel to camera

   Expected: 100% detection, accurate scale

2. Partial A4 Obscuration

   • Object covering corner of A4
   • Object covering 25% of A4
   • Object covering 50% of A4

   Expected: Detection fails above certain threshold Test: Find maximum acceptable obscuration

3. A4 Variations

   • White A4 (standard)
   • Colored A4 (blue, as in WhatTheCameraSees.jpeg)
   • Grid-printed A4
   • Textured A4

   Expected: Color shouldn't matter if high contrast with base Test: Validate --paper-color option

4. Non-A4 Papers

   • A3 paper (297mm × 420mm)
   • A5 paper (148mm × 210mm)
   • Letter paper (216mm × 279mm)

   Expected: Requires --paper-format option Test: Validate other ISO formats

5. No Reference Paper

   • Object alone on base (no A4)
   • Using --reference-mm manual override

   Expected: Falls back to estimation or manual scale Test: Accuracy with manual reference

• Measure A4 detection success rate
• Compare scale accuracy: A4 detected vs. estimated
• Test all supported paper formats
• Document when manual override is necessary

Document system limitations and error handling.

1. Extremely Low Contrast

   • Gray object on gray background
   • White object on white paper

   Expected: Detection fails Validation: Error message is clear

2. Object Color Same as Base Feet

   • Red object on red base feet

   Expected: May confuse detection Test: Base-feet detection impact

3. Highly Irregular Shapes

   • Crumpled paper ball (3D object)
   • Tangled string/wire
   • Natural twig/branch

   Expected: 2D projection may not match expectations Validation: Document as out-of-scope for 3D objects

4. Transparent Objects with Opaque Features

   • Clear ruler with black markings
   • Transparent protractor

   Expected: Only opaque parts detected Test: Partial extraction behavior

5. Motion Blur

   • Intentionally defocused image
   • Camera shake simulation

   Expected: Fuzzy edges degrade quality Test: Quality metric correlation

6. Multiple Overlapping Objects

   • Two objects touching
   • One object partially behind another

   Expected: Detected as single merged object Validation: Document requirement for object separation

• Photograph each edge case
• Document error messages or failure modes
• Categorize: "Works with adjustment", "Acceptable degradation", "Documented limitation"
• Update documentation with limitations

Validate all command-line options work correctly.

1. Color Options (-c, --color)
   • All named colors (black, red, blue, etc.)
   • Hex colors (#FF0000, #00FF00, etc.)
   • Invalid colors (error handling)
2. Margin Options (-m, --margin-mm)
   • 0mm (no margin)
   • 2mm (default)
   • 5mm, 10mm (large margins)
   • Negative value (error handling)
3. Threshold Options (-t, --threshold)
   • Range sweep: 20%, 30%, 40%, 50%, 60%, 70%, 80%
   • Optimal value per object type
   • Auto-threshold (-a) validation
4. Auto-Threshold (-a, --auto-threshold)
   • Different TRIES values (3, 5, 7, 10)
   • Compare auto vs manual results
   • Performance impact measurement
5. No-Holes Flag (--no-holes)
   • Objects with holes (ring, letters O/B/P/A)
   • Objects without holes (should be identical)
   • Known bug validation (if still present)
6. Paper Format (-p, --paper-format)
   • A4, A3, A5, Letter
   • Invalid format (error handling)
7. Debug Mode (-d, --debug)
   • Verify all debug images generated
   • Check debug directory structure
   • Validate debug image content

• Create test matrix: Object × Option × Value
• Automated testing where possible (script-driven)
• Document which options affect which object types
• Create "best practices" guide from results

Validate practical usefulness with actual use-case objects.

1. Office/Craft Supplies

   • Scissors (thin metal)
   • Paperclip (wire)
   • Pen/pencil
   • Stapler
   • USB drive
   • Key

   Why: Common small objects Expected: Mixed results due to reflectivity and thin features

2. Electronic Components

   • Arduino board
   • Raspberry Pi
   • PCB (circuit board)
   • Component footprints

   Why: Technical documentation use case Expected: Good for rectangular boards, challenging for components

3. Natural Objects

   • Leaf (various species)
   • Flower petal
   • Feather
   • Rock/pebble cross-section

   Why: Organic shapes, educational use Expected: Good for flat objects, poor for 3D

4. Fashion/Textiles

   • Button
   • Zipper pull
   • Fabric swatch
   • Pattern template

   Why: Design/craft use case Expected: Flat items good, 3D items poor

5. Maker/Prototyping

   • 3D printed part
   • Laser-cut template
   • Gasket/washer
   • Mechanical part

   Why: Reverse engineering, documentation Expected: Excellent for flat parts

• Prioritize objects from actual user requests/use cases
• Create gallery of successes and failures
• Document workarounds for common object types
• Build "object type → settings" recommendation table

TestInstrumentation/GeometricShapes/ holds one subdirectory per test object, each named for its known dimensions (Circle50mm/, Square25mm/, Rectangle50x100mm/, Triangle80mm/, etc.). The reference calibration target lives separately, in TestInstrumentation/ReferenceCard/ — see ReferenceCard.org.

Each geometric shape directory holds:

1. ObjectName.scad — OpenSCAD source with precise dimensions
2. ObjectName.stl — exported STL for 3D printing (optional, can be regenerated from the .scad)
3. ObjectName.3mf — PrusaSlicer project with the actual manufacturing settings used
4. PhysicalMeasurements.org — caliper measurements after printing
5. Notes.org — observations about the printed object

Org-mode files (points 4-5) were chosen for structured tables and checkboxes, LaTeX/PDF export if a report is needed, and plain-text, version-control-friendly storage — consistent with the rest of the project. .3mf files (point 3) capture the complete manufacturing context (slicer settings, print orientation, material, infill, perimeters, supports) in one place, alongside the .scad source, so a print can be reproduced exactly, what actually worked (vs. what was merely recommended) is on record, and settings are easy to compare if a print fails.

All geometric test shapes were printed with these settings (recorded in each object's own .3mf):

Material

Black PLA (matte finish)

Printer

Prusa XL with 0.6mm HF nozzle

Tool used

Tool 1 / Extruder 1 (second extruder)

Layer height

0.2mm

Infill

100% (solid objects)

Speed

30mm/s (perimeters, infill, external perimeters); 40mm/s first layer

Nozzle temperature

230°C (PLA)

Bed temperature

60°C

30mm/s was chosen deliberately over the printer's 200mm/s capability, trading print time for dimensional accuracy. These are test instrumentation settings, distinct from the hardware/jig components (see ManufacturingConfig.org), which use PETG on tool 0.

Black matte PLA specifically, rather than the jig's PETG: PLA shrinks less than PETG (~0.3% vs ~0.7%, mattering for dimensional accuracy), a matte surface minimizes reflections in photographs, black gives high contrast against white A4 paper, PLA prints at lower temperature with less warping, and it is lower cost for consumable, deliberately-replaceable test objects.

After printing:

1. Remove carefully from the bed; let cool completely (30+ minutes)
2. Remove any brim/raft carefully
3. Measure with calipers (±0.1mm precision)
4. Record actual dimensions in PhysicalMeasurements.org
5. Compare to the .scad source's nominal dimensions
6. Accept if within ±0.2mm tolerance

The known dimensions — from both the .scad source and the physical caliper measurements — are what dvi3v's extracted SVG dimensions are validated against.