{ "testId": "SAIL-165", "testNumber": "3.1", "name": "model-yolo26-swap", "phase": "PRE-RELABEL BASELINE", "generatedAt": "2026-09-07T14:13:29.492394+00:00", "productQuestion": "Does swapping our finder to YOLO26-pose find more closely stacked and faint bottom stripes?", "status": "blocked_data_integrity", "productVerdict": "indeterminate", "bottomLine": "Unknown: no demonstrated improvement for a sailor. YOLO26 training and paired inference did not run; the delta versus current SailScan v26 is unmeasured.", "whatItMeansForSailScan": "The NMS-free hypothesis remains untested. The current input snapshot contains a test fixture instead of a photo and unresolved image/label frame conflicts. Establish the correct frames and source image before measuring the finder swap; otherwise a recall difference could be a data error.", "holdout": { "file": "server/inference/test-holdout-ids-canonical.json", "frozen": "2026-08-25", "count": 332, "sha256": "a49d9139d6eac54951c4c3e257770855192a75f8c941799ef4141c49a4c11243", "idsChanged": false, "trainOverlap": 0, "validationOverlap": 0, "missingIds": [], "unresolvedFrameIds": [ "68ccaf678f1d2c0cda50c1a7", "69c44d5c8c718d5d543a2010", "69cea47eed067217da126fc4" ] }, "execution": { "cloudAccessAvailable": true, "podAllocated": true, "podId": "ubgu8igk4t349a", "gpuRequested": "NVIDIA RTX A5000", "advertisedUsdPerHour": 0.27, "podTerminated": true, "gpuTrainingStarted": false, "completedEpochs": 0, "candidateWeightsProduced": false, "newBaselinePhotosInferred": 0, "candidatePhotosInferred": 0, "localHeavyTraining": false, "setupExecutedOnPod": false, "whyStopped": "CPU preflight failed before training: unreadable retained image and unresolved EXIF frame provenance. This was not a RunPod capacity/authentication failure. Pod was still initializing when terminated.", "notRun": [ "GPU venv installation and model-head assertion", "YOLO dataset preparation", "YOLO26 training", "paired incumbent/candidate holdout inference", "full production pipeline comparison", "faint/stacked slice evaluation" ] }, "inputAudit": { "status": "blocked_data_integrity", "inputRecords": 4956, "gatedRecords": 4323, "metadataSplitCounts": { "train": 3421, "val": 538, "test": 364 }, "holdoutFile": "test-holdout-ids-canonical.json", "holdoutFrozen": "2026-08-25", "holdoutSha256": "a49d9139d6eac54951c4c3e257770855192a75f8c941799ef4141c49a4c11243", "canonicalHoldout": 332, "missingHoldout": [], "leakedHoldout": [], "invalidImages": [ { "id": "35e717a7-ca96-464b-a424-d609ca04ea1d", "split": "train", "canonicalHoldout": false, "error": "cannot identify image file '/Volumes/MacMiniExtra/Projects/boats/server/inference/training-data-mixed/images/app__35e717a7-ca96-464b-a424-d609ca04ea1d.jpg'", "bytes": 24 } ], "frameConflicts": [ { "id": "68bf318c58038aac44157e21", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 1600, 2133 ], "rawSize": [ 1600, 2133 ], "uprightSize": [ 2133, 1600 ] }, { "id": "68ccaf678f1d2c0cda50c1a7", "split": "test", "canonicalHoldout": true, "orientation": 6, "labelSize": [ 1600, 1200 ], "rawSize": [ 1600, 1200 ], "uprightSize": [ 1200, 1600 ] }, { "id": "69052d835821020fc4dbf748", "split": "val", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1600, 2133 ], "rawSize": [ 1600, 2133 ], "uprightSize": [ 2133, 1600 ] }, { "id": "69127135e05a8133399d0283", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1440, 1920 ], "rawSize": [ 1440, 1920 ], "uprightSize": [ 1920, 1440 ] }, { "id": "6927371a99113bd5e245012c", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 1600, 1200 ], "rawSize": [ 1600, 1200 ], "uprightSize": [ 1200, 1600 ] }, { "id": "692d58d73334ff75ab7b2791", "split": "val", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 1600, 1200 ], "rawSize": [ 1600, 1200 ], "uprightSize": [ 1200, 1600 ] }, { "id": "6970988d3763c4e1ac2e11cd", "split": "train", "canonicalHoldout": false, "orientation": 5, "labelSize": [ 1200, 1599 ], "rawSize": [ 1200, 1599 ], "uprightSize": [ 1599, 1200 ] }, { "id": "6983d41a3b15d0ca7a33597a", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1600, 2133 ], "rawSize": [ 1600, 2133 ], "uprightSize": [ 2133, 1600 ] }, { "id": "698da59b495f31e65b068ab6", "split": "train", "canonicalHoldout": false, "orientation": 5, "labelSize": [ 1200, 1599 ], "rawSize": [ 1200, 1599 ], "uprightSize": [ 1599, 1200 ] }, { "id": "698dbf7a495f31e65b068ab9", "split": "train", "canonicalHoldout": false, "orientation": 5, "labelSize": [ 1200, 1599 ], "rawSize": [ 1200, 1599 ], "uprightSize": [ 1599, 1200 ] }, { "id": "6994b0cf66293f7fd25a5141", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 600, 800 ], "rawSize": [ 600, 800 ], "uprightSize": [ 800, 600 ] }, { "id": "69970f436a4cc508426dbba6", "split": "val", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1440, 1920 ], "rawSize": [ 1440, 1920 ], "uprightSize": [ 1920, 1440 ] }, { "id": "69b7f2a7cb608188571566b7", "split": "train", "canonicalHoldout": false, "orientation": 5, "labelSize": [ 1200, 1599 ], "rawSize": [ 1200, 1599 ], "uprightSize": [ 1599, 1200 ] }, { "id": "69bc2539f447c9b658333c9f", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 1600, 2133 ], "rawSize": [ 1600, 2133 ], "uprightSize": [ 2133, 1600 ] }, { "id": "69bc263ef447c9b658333ca0", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 1600, 2133 ], "rawSize": [ 1600, 2133 ], "uprightSize": [ 2133, 1600 ] }, { "id": "69c2efc6613f65a9f096f91d", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1536, 2048 ], "rawSize": [ 1536, 2048 ], "uprightSize": [ 2048, 1536 ] }, { "id": "69c44c908c718d5d543a200f", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1600, 2844 ], "rawSize": [ 1600, 2844 ], "uprightSize": [ 2844, 1600 ] }, { "id": "69c44d5c8c718d5d543a2010", "split": "test", "canonicalHoldout": true, "orientation": 8, "labelSize": [ 1600, 2844 ], "rawSize": [ 1600, 2844 ], "uprightSize": [ 2844, 1600 ] }, { "id": "69c44ff8873550ef9299bea6", "split": "val", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1600, 2844 ], "rawSize": [ 1600, 2844 ], "uprightSize": [ 2844, 1600 ] }, { "id": "69c4511e407e8b260369f4a7", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 1600, 2844 ], "rawSize": [ 1600, 2844 ], "uprightSize": [ 2844, 1600 ] }, { "id": "69cc2cf6b808bf7afc49fbf8", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 960, 1280 ], "rawSize": [ 960, 1280 ], "uprightSize": [ 1280, 960 ] }, { "id": "69cea47eed067217da126fc4", "split": "test", "canonicalHoldout": true, "orientation": 8, "labelSize": [ 1600, 2133 ], "rawSize": [ 1600, 2133 ], "uprightSize": [ 2133, 1600 ] }, { "id": "69d545ad7714997a7adce6b6", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 1536, 2048 ], "rawSize": [ 1536, 2048 ], "uprightSize": [ 2048, 1536 ] }, { "id": "69e83c0bf01f76b6242ecfbe", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1440, 1920 ], "rawSize": [ 1440, 1920 ], "uprightSize": [ 1920, 1440 ] }, { "id": "69ea5a9eda0bdee6ef4a2ad0", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1536, 2048 ], "rawSize": [ 1536, 2048 ], "uprightSize": [ 2048, 1536 ] }, { "id": "69ea5bbcda0bdee6ef4a2ad1", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1536, 2048 ], "rawSize": [ 1536, 2048 ], "uprightSize": [ 2048, 1536 ] }, { "id": "69f8cea4ea23d5ac8b291715", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1536, 2048 ], "rawSize": [ 1536, 2048 ], "uprightSize": [ 2048, 1536 ] }, { "id": "69fe3c90a21d8b7ab38d0bd2", "split": "train", "canonicalHoldout": false, "orientation": 6, "labelSize": [ 1600, 2133 ], "rawSize": [ 1600, 2133 ], "uprightSize": [ 2133, 1600 ] }, { "id": "6a1dea4c4d15a0ffad42f713", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1200, 1600 ], "rawSize": [ 1200, 1600 ], "uprightSize": [ 1600, 1200 ] }, { "id": "6a1dec2f4d15a0ffad42f714", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1200, 1600 ], "rawSize": [ 1200, 1600 ], "uprightSize": [ 1600, 1200 ] }, { "id": "6a1ded384d15a0ffad42f715", "split": "val", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1200, 1600 ], "rawSize": [ 1200, 1600 ], "uprightSize": [ 1600, 1200 ] }, { "id": "6a2aadf2f24e2bd18f3d451b", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1536, 2048 ], "rawSize": [ 1536, 2048 ], "uprightSize": [ 2048, 1536 ] }, { "id": "6a392e0b8c6926f954e579ff", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 1200, 1600 ], "rawSize": [ 1200, 1600 ], "uprightSize": [ 1600, 1200 ] }, { "id": "6a43b7c25d7865a23e87f2b2", "split": "train", "canonicalHoldout": false, "orientation": 8, "labelSize": [ 627, 835 ], "rawSize": [ 627, 835 ], "uprightSize": [ 835, 627 ] } ], "note": "Metadata split audit only. Not YOLO loader counts. Frame conflicts require provenance review; dimension evidence does not authorize rotating labels. Test split includes additional test-only records/crop siblings; evaluation remains exactly canonical 332." }, "metrics": { "candidateBottomRecall": null, "bottomRecallDeltaPercentagePoints": null, "candidateFaintRecall": null, "faintRecallDeltaPercentagePoints": null, "candidateStackedRecall": null, "stackedRecallDeltaPercentagePoints": null, "candidatePrecision": null, "additionalFalsePositives": null, "betterForSailor": null }, "historicalIncumbentContext": { "source": "server/inference/runs/prod-baseline-20260825/report.json", "sha256": "538be36469f87621c0028c6420b8017ce1eee720cb3f89c1a09d70f445c5d541", "modelVersion": "sailscan-v26", "architecture": "YOLOv8m-pose", "holdoutFile": "server/inference/test-holdout-ids-canonical.json", "holdoutFrozen": "2026-08-25", "sourceReportedValidity": { "status": "complete", "scored": 332, "expected": 332, "skipped": 0, "modelVersions": [ "sailscan-v26" ], "note": "complete single-version run" }, "bottomMatched": 300, "bottomTotal": 332, "bottomRecall": 0.9036144578313253, "bottomRecallPercent": 90.36144578313252, "bottomMissed": 32, "linesPerfect": 259, "linesPerfectTotal": 332, "linesPerfectPercent": 78.01204819277109, "newMeasurement": false, "caveat": "Historical cached report only; not rerun or newly frame-validated. Three canonical photos have unresolved frame provenance in the present local snapshot. This is not a matched YOLO26 comparison." }, "runSetup": { "candidate": "yolo26m-pose.pt", "incumbent": "models/sailscan-v26-curvature.pt", "ultralyticsPin": "8.4.142", "incumbentTrainingUltralyticsPin": "8.4.92", "torchPin": "2.8.0", "torchvisionPin": "0.23.0", "epochsPlanned": 100, "batch": 8, "imgsz": 960, "keypointSampling": "curvature", "keypointsPerStripe": 8, "seed": 0, "degrees": 30, "flipud": 0.3, "erasing": 0.4, "hsvV": 0.6, "labelsSnapshot": "training-data-mixed/labels.jsonl", "inputHashes": { "train.py": "a6cd77fe9a89dbd9d7ca141b04d68e06f54cb1e91e22c809ec91cd003ab5f5a9", "test-holdout-ids-canonical.json": "a49d9139d6eac54951c4c3e257770855192a75f8c941799ef4141c49a4c11243", "training-data-mixed/labels.jsonl": "39a276e6256a8d28e4bcb6ae782592291423203dec7abbf26a8319dfb7e71441", "excluded.json": "e90e20579b69f0891846bfd2ed22c5c910c0010f182492c701474e47cf5a1784", "models/sailscan-v26-curvature.pt": "0086b5d5caa7b6001408d355f21efa50a22048e836562456ac2a7839534689ae", "prod_baseline_score.mjs": "3f7a864e935cb562dd12e56a7a109dd0353d1d2202ad65899bc2a8f5b46601fa" }, "launcher": "server/inference/experiments/sail165-yolo26/run-on-pod.sh", "protocol": "server/inference/experiments/sail165-yolo26/README.md", "preflight": "server/inference/experiments/sail165-yolo26/preflight.py" }, "measurementLimitations": [ "No yes/no result about NMS-free stripe separation is supported. Null means unmeasured, not zero improvement.", "Gate 0 uncapped greedy recall is not spatially gated recall; use the existing holdout suite optimal matching and 1/2/3/5%-diagonal cap sweep.", "Faintness and closely stacked slices were not human-tagged for this experiment; do not substitute low confidence for faintness.", "Swapping architecture also changes heads/losses; comparing newly trained current labels with an older incumbent does not isolate the causal effect of NMS.", "Metadata split counts (3421 train, 538 validation, 364 test-only records) are pre-loader counts, not completed training counts. The canonical scoring denominator remains 332." ], "guardrails": { "holdoutNeverTrained": true, "weightsPromoted": false, "cloudflareDeployed": false, "indexHtmlEdited": false, "gitCommitted": false }, "sources": [ "https://docs.ultralytics.com/models/yolo26/", "https://docs.ultralytics.com/guides/end2end-detection/", "https://pypi.org/project/ultralytics/8.4.142/" ] }