Dataset Viewer
Auto-converted to Parquet Duplicate
text
string
33 0.394258 0.415370 0.009349 0.028796
43 0.491406 0.712176 0.007448 0.029259
33 0.417878 0.401574 0.010130 0.028519
43 0.516419 0.711736 0.007318 0.029676
50 0.323099 0.479468 0.012656 0.024306
42 0.343008 0.520185 0.013411 0.024444
50 0.305130 0.508889 0.012604 0.023519
2 0.382422 0.468333 0.011458 0.028796
2 0.395885 0.518958 0.011354 0.028380
43 0.516784 0.764028 0.007422 0.028889
43 0.491745 0.764468 0.007344 0.028565
2 0.418359 0.500949 0.011875 0.028380
42 0.307604 0.550648 0.010260 0.026852
42 0.324453 0.596667 0.010677 0.025648
42 0.344180 0.571713 0.010755 0.026667
42 0.327500 0.526134 0.010990 0.027546
2 0.404792 0.450370 0.011146 0.028704
5 0.283138 0.619653 0.021276 0.031065
28 0.275312 0.669352 0.014531 0.020556
28 0.251107 0.638449 0.014141 0.023843
28 0.288659 0.632986 0.014036 0.022824
28 0.264453 0.601181 0.014792 0.021806
50 0.492096 0.694745 0.008464 0.028565
50 0.510638 0.633704 0.008828 0.028704
50 0.486237 0.642431 0.008464 0.029120
50 0.516263 0.685949 0.008932 0.028843
31 0.388516 0.448773 0.011198 0.029861
31 0.400508 0.497894 0.011901 0.029491
43 0.507799 0.399630 0.008464 0.029907
41 0.331680 0.499954 0.011484 0.025648
43 0.483177 0.405833 0.007760 0.029352
10 0.587682 0.413426 0.009323 0.028796
18 0.679818 0.489537 0.012031 0.025463
10 0.611081 0.425926 0.009193 0.028611
43 0.487227 0.459560 0.008620 0.028102
31 0.422995 0.480509 0.011094 0.028148
18 0.697878 0.516944 0.011693 0.025000
10 0.601771 0.478171 0.008958 0.028843
40 0.350143 0.542431 0.011484 0.024120
43 0.512135 0.455671 0.007396 0.023102
4 0.301081 0.669514 0.014609 0.019306
31 0.411172 0.431042 0.011771 0.029306
10 0.578503 0.465417 0.008932 0.028889
12 0.748867 0.638495 0.013203 0.021806
12 0.728854 0.675046 0.013021 0.021296
4 0.312513 0.630671 0.014714 0.021343
12 0.712279 0.645208 0.012891 0.022639
12 0.732760 0.609375 0.012917 0.022083
18 0.659753 0.532546 0.012057 0.024815
40 0.317734 0.570370 0.011094 0.027037
40 0.355430 0.583542 0.011172 0.026157
40 0.336549 0.612407 0.012526 0.026389
28 0.250964 0.638333 0.014010 0.023981
18 0.678216 0.559884 0.012318 0.025417
28 0.275273 0.669398 0.014349 0.020370
40 0.334974 0.540949 0.009115 0.025880
28 0.286719 0.603287 0.015208 0.020556
4 0.264336 0.601458 0.014922 0.021713
4 0.274987 0.641319 0.014297 0.022269
27 0.404661 0.542616 0.012500 0.029861
45 0.498620 0.503403 0.008802 0.029306
19 0.643893 0.504722 0.011849 0.028704
45 0.474466 0.512454 0.010026 0.029815
27 0.427760 0.526875 0.011667 0.028287
31 0.388464 0.448843 0.011250 0.030000
50 0.454857 0.711157 0.007786 0.028889
50 0.479622 0.711574 0.007891 0.028519
50 0.479440 0.764167 0.007734 0.028611
50 0.454779 0.763657 0.007266 0.028981
45 0.467669 0.460394 0.008672 0.029213
47 0.482969 0.405833 0.007865 0.029815
15 0.303464 0.713241 0.013594 0.020833
9 0.575937 0.507454 0.008854 0.029722
40 0.331667 0.499907 0.011458 0.025741
10 0.587708 0.413449 0.009010 0.028935
10 0.611042 0.425856 0.008958 0.028750
43 0.507747 0.399421 0.008464 0.030046
15 0.279466 0.685463 0.014401 0.021852
15 0.315312 0.677546 0.014219 0.020648
43 0.512109 0.452014 0.007917 0.029398
18 0.679818 0.489329 0.012031 0.025602
10 0.601888 0.478287 0.009193 0.029167
9 0.551706 0.498912 0.008568 0.029120
31 0.411081 0.430972 0.011953 0.029352
18 0.697878 0.516991 0.011589 0.025093
19 0.649909 0.573426 0.012214 0.027870
9 0.558125 0.445718 0.008906 0.029583
41 0.684180 0.645486 0.013307 0.022176
31 0.393203 0.491782 0.011094 0.029861
19 0.628737 0.552431 0.011589 0.026991
12 0.748672 0.638935 0.013802 0.021852
19 0.665000 0.525949 0.010937 0.028102
15 0.291016 0.648889 0.013177 0.020926
41 0.312643 0.527523 0.010495 0.025046
40 0.350078 0.542315 0.011250 0.024259
41 0.701549 0.673079 0.013151 0.022639
18 0.678451 0.559954 0.012161 0.025000
12 0.728737 0.675116 0.012422 0.021157
40 0.355000 0.583056 0.011563 0.026667
4 0.312227 0.630417 0.014974 0.022130
End of preview. Expand in Data Studio

LGD Cards — Frozen Video Holdout (evaluation set)

The frozen holdout used to score every generation of the Live Game Defender (LGD) card detector. Never trained on — this is the reproducibility benchmark behind the recall numbers on lgd-cards-gen1/2/3/4.

  • Format: YOLO — images/ + labels/ (full frames, not tiles), data.yaml (52 classes, eval only).
  • 318 frames, ~6,713 GT pip boxes, drawn from held-out PoC recordings. Grew 225 → 318 on 2026-07-20: +93 held-out deck-spread frames added for gen4 (KAS-52 spread-recall campaign).
  • Scoring: serve-style auto-grid tiling, threshold 50, code-aware (recall + precision proxy). Trend on the historical 225-frame suite: stock 0.539 → gen2 0.680 → gen3 0.847. On this expanded 318-frame suite: gen3 0.750 → gen4 0.786 (and on the 93 spread-only frames, 0.574 → 0.665). Compare models only within one suite.

⚠️ Honesty note

Our own PoC / academy footage, ground truth LLM-verified with human corrections — a consistent internal benchmark, not a validated real-world casino-accuracy dataset (rule #5 of the project). Use it to compare models against each other, not to claim floor accuracy.

Family

Datasets: day-1 · day-2 · day-3 · holdout (this). Models: lgd-cards-gen1/2/3.

License CC-BY-NC-4.0 (non-commercial, attribution). © 2026 TechTools s.r.o.

Downloads last month
57