From 22192ec1a33024c6c8f1845bea621e69811fcc46 Mon Sep 17 00:00:00 2001
From: jungfrau <jungfrau@mx-spark-01.psi.ch>
Date: Mon, 26 Oct 2026 09:16:33 +0100
Subject: [PATCH] run 2026-10-26_v5h-noflip-mask1: gate_failed

---
 .../gold_report.json                          | 118 ++++++++++++++++++
 .../metadata.json                             |  71 +++++++++++
 .../weights.pt                                |   3 +
 runs/2026-10-26_v5h-noflip-mask1/latency.json |   1 +
 runs/2026-10-26_v5h-noflip-mask1/metrics.json |  57 +++++++++
 runs/2026-10-26_v5h-noflip-mask1/run.json     |  52 +++++++-
 runs/README.md                                |   8 ++
 7 files changed, 305 insertions(+), 5 deletions(-)
 create mode 100644 models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/gold_report.json
 create mode 100644 models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/metadata.json
 create mode 100644 models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/weights.pt
 create mode 100644 runs/2026-10-26_v5h-noflip-mask1/latency.json
 create mode 100644 runs/2026-10-26_v5h-noflip-mask1/metrics.json

diff --git a/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/gold_report.json b/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/gold_report.json
new file mode 100644
index 00000000..91bda2bd
--- /dev/null
+++ b/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/gold_report.json
@@ -0,0 +1,118 @@
+{
+  "per_class": {
+    "loop_all": {
+      "tp": 147,
+      "fp": 32,
+      "fn": 11,
+      "precision": 0.8212290502793296,
+      "recall": 0.930379746835443,
+      "f1": 0.8724035608308606,
+      "mean_mask_iou": 0.8915905247976269,
+      "n_matched": 147
+    },
+    "pin": {
+      "tp": 74,
+      "fp": 23,
+      "fn": 10,
+      "precision": 0.7628865979381443,
+      "recall": 0.8809523809523809,
+      "f1": 0.8176795580110496,
+      "mean_mask_iou": 0.915598502368835,
+      "n_matched": 74
+    },
+    "crystal": {
+      "tp": 80,
+      "fp": 66,
+      "fn": 55,
+      "precision": 0.547945205479452,
+      "recall": 0.5925925925925926,
+      "f1": 0.5693950177935944,
+      "mean_mask_iou": 0.7838964045966136,
+      "n_matched": 80
+    },
+    "loop_face": {
+      "tp": 129,
+      "fp": 68,
+      "fn": 26,
+      "precision": 0.6548223350253807,
+      "recall": 0.832258064516129,
+      "f1": 0.7329545454545453,
+      "mean_mask_iou": 0.8581593922332215,
+      "n_matched": 129
+    },
+    "ice": {
+      "tp": 2,
+      "fp": 11,
+      "fn": 40,
+      "precision": 0.15384615384615385,
+      "recall": 0.047619047619047616,
+      "f1": 0.07272727272727272,
+      "mean_mask_iou": 0.8541725321804396,
+      "n_matched": 2
+    },
+    "needle": {
+      "tp": 0,
+      "fp": 0,
+      "fn": 1,
+      "precision": 0.0,
+      "recall": 0.0,
+      "f1": 0.0,
+      "mean_mask_iou": 0.0,
+      "n_matched": 0
+    }
+  },
+  "overall": {
+    "tp": 432,
+    "fp": 200,
+    "fn": 143,
+    "precision": 0.6835443037974683,
+    "recall": 0.7513043478260869,
+    "f1": 0.715824357912179,
+    "mean_mask_iou": 0.8656035077562974,
+    "n_matched": 432
+  },
+  "dataset": {
+    "gold_images": 335,
+    "pred_images": 335,
+    "missing_predictions": 0,
+    "iou_threshold": 0.5,
+    "ignored_objects": 454
+  },
+  "environment": {
+    "host": "mx-spark-01",
+    "ultralytics": "8.4.160",
+    "torch": "2.14.0+cu130"
+  },
+  "gate": {
+    "baseline": "/home/jungfrau/AareLC_train/models/registry/yolo26s-seg-v5-s-noflip-mask1-2026-10-25/gold_report.json",
+    "f1_delta": -0.0448,
+    "mask_iou_delta": -0.0327,
+    "mask_score": 0.6503,
+    "mask_score_baseline": 0.6831,
+    "matched_only_iou_delta": 0.0043,
+    "class_f1_delta": {
+      "loop_all": -0.0483,
+      "pin": -0.0221
+    },
+    "pass_point_rule": false,
+    "rule": "paired bootstrap",
+    "bootstrap": {
+      "frames": 335,
+      "mounts": 300,
+      "draws": 2000,
+      "f1": {
+        "delta": -0.0448,
+        "lo": -0.074,
+        "hi": -0.0181,
+        "win": 0.001
+      },
+      "mask_score": {
+        "delta": -0.0327,
+        "lo": -0.0544,
+        "hi": -0.0126,
+        "win": 0.001
+      }
+    },
+    "pass": false
+  }
+}
diff --git a/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/metadata.json b/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/metadata.json
new file mode 100644
index 00000000..4a123094
--- /dev/null
+++ b/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/metadata.json
@@ -0,0 +1,71 @@
+{
+  "model_id": "yolo26n-seg-v5h-noflip-mask1-2026-10-26",
+  "family": "yolo26n",
+  "task": "segmentation",
+  "training_date": "2026-10-26",
+  "run": "2026-10-26_v5h-noflip-mask1",
+  "dataset": "dataset/2026-10-26_v5h",
+  "base_weights": "models/base/yolo26n-seg.pt",
+  "hyperparams": {
+    "epochs": 400,
+    "patience": 50,
+    "imgsz": 640,
+    "batch": 16,
+    "device": "0",
+    "overlap_mask": false,
+    "scale": 0.7,
+    "translate": 0.15,
+    "mosaic": 0.5,
+    "close_mosaic": 10,
+    "fliplr": 0,
+    "flipud": 0.3,
+    "shear": 5.0,
+    "perspective": 0.0,
+    "mask_ratio": 1
+  },
+  "export_formats": [
+    "pt"
+  ],
+  "metrics": {
+    "tp": 432,
+    "fp": 200,
+    "fn": 143,
+    "precision": 0.6835443037974683,
+    "recall": 0.7513043478260869,
+    "f1": 0.715824357912179,
+    "mean_mask_iou": 0.8656035077562974,
+    "n_matched": 432
+  },
+  "gate": {
+    "baseline": "/home/jungfrau/AareLC_train/models/registry/yolo26s-seg-v5-s-noflip-mask1-2026-10-25/gold_report.json",
+    "f1_delta": -0.0448,
+    "mask_iou_delta": -0.0327,
+    "mask_score": 0.6503,
+    "mask_score_baseline": 0.6831,
+    "matched_only_iou_delta": 0.0043,
+    "class_f1_delta": {
+      "loop_all": -0.0483,
+      "pin": -0.0221
+    },
+    "pass_point_rule": false,
+    "rule": "paired bootstrap",
+    "bootstrap": {
+      "frames": 335,
+      "mounts": 300,
+      "draws": 2000,
+      "f1": {
+        "delta": -0.0448,
+        "lo": -0.074,
+        "hi": -0.0181,
+        "win": 0.001
+      },
+      "mask_score": {
+        "delta": -0.0327,
+        "lo": -0.0544,
+        "hi": -0.0126,
+        "win": 0.001
+      }
+    },
+    "pass": false
+  }
+}
diff --git a/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/weights.pt b/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/weights.pt
new file mode 100644
index 00000000..d54ba9b8
--- /dev/null
+++ b/models/registry/yolo26n-seg-v5h-noflip-mask1-2026-10-26/weights.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f60c36a7cca90765a1e8393ca30e38f14de6a192bd0ec43c949232d64ce9e4a7
+size 6581277
diff --git a/runs/2026-10-26_v5h-noflip-mask1/latency.json b/runs/2026-10-26_v5h-noflip-mask1/latency.json
new file mode 100644
index 00000000..412c1468
--- /dev/null
+++ b/runs/2026-10-26_v5h-noflip-mask1/latency.json
@@ -0,0 +1 @@
+{"weights": "runs/2026-10-26_v5h-noflip-mask1/ultralytics/train/weights/best.pt", "frames": 150, "imgsz": 640, "infer_scale": 0.5, "half": true, "ms_median": 4.23, "ms_p90": 4.35, "ms_max": 7.1, "fps_median": 236.2, "device": "0", "backend": "pytorch", "gpu": "NVIDIA GB10", "ultralytics": "8.4.160", "torch": "2.14.0+cu130"}
diff --git a/runs/2026-10-26_v5h-noflip-mask1/metrics.json b/runs/2026-10-26_v5h-noflip-mask1/metrics.json
new file mode 100644
index 00000000..53618a8e
--- /dev/null
+++ b/runs/2026-10-26_v5h-noflip-mask1/metrics.json
@@ -0,0 +1,57 @@
+{
+  "overall": {
+    "tp": 432,
+    "fp": 200,
+    "fn": 143,
+    "precision": 0.6835443037974683,
+    "recall": 0.7513043478260869,
+    "f1": 0.715824357912179,
+    "mean_mask_iou": 0.8656035077562974,
+    "n_matched": 432
+  },
+  "gate": {
+    "baseline": "/home/jungfrau/AareLC_train/models/registry/yolo26s-seg-v5-s-noflip-mask1-2026-10-25/gold_report.json",
+    "f1_delta": -0.0448,
+    "mask_iou_delta": -0.0327,
+    "mask_score": 0.6503,
+    "mask_score_baseline": 0.6831,
+    "matched_only_iou_delta": 0.0043,
+    "class_f1_delta": {
+      "loop_all": -0.0483,
+      "pin": -0.0221
+    },
+    "pass_point_rule": false,
+    "rule": "paired bootstrap",
+    "bootstrap": {
+      "frames": 335,
+      "mounts": 300,
+      "draws": 2000,
+      "f1": {
+        "delta": -0.0448,
+        "lo": -0.074,
+        "hi": -0.0181,
+        "win": 0.001
+      },
+      "mask_score": {
+        "delta": -0.0327,
+        "lo": -0.0544,
+        "hi": -0.0126,
+        "win": 0.001
+      }
+    },
+    "pass": false
+  },
+  "training": {
+    "epochs_run": 188,
+    "precision(B)": 0.72206,
+    "recall(B)": 0.72365,
+    "mAP50(B)": 0.71787,
+    "mAP50-95(B)": 0.57152,
+    "precision(M)": 0.71814,
+    "recall(M)": 0.71711,
+    "mAP50(M)": 0.70593,
+    "mAP50-95(M)": 0.55179
+  },
+  "version": "yolo26n-seg-v5h-noflip-mask1-2026-10-26",
+  "measured_at": "2026-10-26T09:16:24"
+}
diff --git a/runs/2026-10-26_v5h-noflip-mask1/run.json b/runs/2026-10-26_v5h-noflip-mask1/run.json
index b31e0e37..c8f665ab 100644
--- a/runs/2026-10-26_v5h-noflip-mask1/run.json
+++ b/runs/2026-10-26_v5h-noflip-mask1/run.json
@@ -2,18 +2,60 @@
   "id": "2026-10-26_v5h-noflip-mask1",
   "dataset": "dataset/2026-10-26_v5h",
   "base_weights": "models/base/yolo26n-seg.pt",
-  "status": "exported",
+  "status": "gate_failed",
   "created": "2026-10-26T01:43:39",
   "created_by": "jungfrau@mx-spark-01",
-  "hyperparams": {},
-  "version": null,
-  "metrics": {},
+  "hyperparams": {
+    "epochs": 400,
+    "patience": 50,
+    "imgsz": 640,
+    "batch": 16,
+    "device": "0",
+    "overlap_mask": false,
+    "scale": 0.7,
+    "translate": 0.15,
+    "mosaic": 0.5,
+    "close_mosaic": 10,
+    "fliplr": 0,
+    "flipud": 0.3,
+    "shear": 5.0,
+    "perspective": 0.0,
+    "mask_ratio": 1
+  },
+  "version": "yolo26n-seg-v5h-noflip-mask1-2026-10-26",
+  "metrics": {
+    "f1": 0.715824357912179,
+    "mean_mask_iou": 0.8656035077562974,
+    "gate_pass": false,
+    "baseline": "/home/jungfrau/AareLC_train/models/registry/yolo26s-seg-v5-s-noflip-mask1-2026-10-25/gold_report.json",
+    "measured_at": "2026-10-26T09:16:24"
+  },
   "history": [
     {
       "status": "exported",
       "at": "2026-10-26T01:43:39",
       "by": "jungfrau@mx-spark-01",
       "note": "promoted recipe (fliplr 0, mask_ratio 1) on dataset v5h: v5 with half of the accepted-unchanged frames (export.accepted_share 0.5, rare classes kept), gate only, Spark"
+    },
+    {
+      "status": "training",
+      "at": "2026-10-26T01:43:39",
+      "by": "jungfrau@mx-spark-01",
+      "note": "train_run.py"
+    },
+    {
+      "status": "measured",
+      "at": "2026-10-26T09:16:24",
+      "by": "jungfrau@mx-spark-01",
+      "note": ""
+    },
+    {
+      "status": "gate_failed",
+      "at": "2026-10-26T09:16:24",
+      "by": "jungfrau@mx-spark-01",
+      "note": "{\"baseline\": \"/home/jungfrau/AareLC_train/models/registry/yolo26s-seg-v5-s-noflip-mask1-2026-10-25/gold_report.json\", \"f1_delta\": -0.0448, \"mask_iou_delta\": -0.0327, \"mask_score\": 0.6503, \"mask_score_baseline\": 0.6831, \"matched_only_iou_delta\": 0.0043, \"class_f1_delta\": {\"loop_all\": -0.0483, \"pin\": -0.0221}, \"pass_point_rule\": false, \"rule\": \"paired bootstrap\", \"bootstrap\": {\"frames\": 335, \"mounts\": 300, \"draws\": 2000, \"f1\": {\"delta\": -0.0448, \"lo\": -0.074, \"hi\": -0.0181, \"win\": 0.001}, \"mask_score\": {\"delta\": -0.0327, \"lo\": -0.0544, \"hi\": -0.0126, \"win\": 0.001}}, \"pass\": false}"
     }
-  ]
+  ],
+  "host": "mx-spark-01",
+  "started": "2026-10-26T01:43:39"
 }
diff --git a/runs/README.md b/runs/README.md
index 70a442af..39d7be72 100644
--- a/runs/README.md
+++ b/runs/README.md
@@ -26,6 +26,7 @@ Generated by `python tools/ledger.py render`; do not edit by hand.
 | 2026-10-23_v5-s-noflip-mask1 | promoted | dataset/2026-10-22_v5 | models/base/yolo26s-seg.pt | yolo26s-seg-v5-s-noflip-mask1-2026-10-25 | 0.761 | 0.861 | yes | 2026-10-25T22:15 |
 | 2026-10-25_v5-amp | gate_failed | dataset/2026-10-22_v5 | models/base/yolo26n-seg.pt | yolo26n-seg-v5-amp-2026-10-25 | 0.746 | 0.865 | no | 2026-10-25T21:37 |
 | 2026-10-25_v5-warm | gate_failed | dataset/2026-10-22_v5 | models/registry/yolo26n-seg-v4-noflip-mask1-2026-10-02/weights.pt | weights-seg-v5-warm-2026-10-26 | 0.757 | 0.861 | no | 2026-10-26T01:43 |
+| 2026-10-26_v5h-noflip-mask1 | gate_failed | dataset/2026-10-26_v5h | models/base/yolo26n-seg.pt | yolo26n-seg-v5h-noflip-mask1-2026-10-26 | 0.716 | 0.866 | no | 2026-10-26T09:16 |
 
 States: exported → queued → training → measured → gate_passed → gate_failed → promoted → deployed (or aborted).
 
@@ -212,3 +213,10 @@ States: exported → queued → training → measured → gate_passed → gate_f
 - 2026-10-25T21:54:51 · training · jungfrau@mx-spark-01 — train_run.py
 - 2026-10-26T01:43:22 · measured · jungfrau@mx-spark-01
 - 2026-10-26T01:43:22 · gate_failed · jungfrau@mx-spark-01 — {"baseline": "/home/jungfrau/AareLC_train/models/registry/yolo26s-seg-v5-s-noflip-mask1-2026-10-25/gold_report.json", "f1_delta": -0.0039, "mask_iou_delta": -0.0075, "mask_score": 0.6755, "mask_score_baseline": 0.6831, "matched_only_iou_delta": -0.0001, "class_f1_delta": {"loop_all": 0.001, "pin": -0.0102}, "pass_point_rule": false, "rule": "paired bootstrap", "bootstrap": {"frames": 335, "mounts": 300, "draws": 2000, "f1": {"delta": -0.0039, "lo": -0.0266, "hi": 0.0186, "win": 0.377}, "mask_score": {"delta": -0.0075, "lo": -0.0241, "hi": 0.0084, "win": 0.171}}, "pass": false}
+
+## 2026-10-26_v5h-noflip-mask1
+
+- 2026-10-26T01:43:39 · exported · jungfrau@mx-spark-01 — promoted recipe (fliplr 0, mask_ratio 1) on dataset v5h: v5 with half of the accepted-unchanged frames (export.accepted_share 0.5, rare classes kept), gate only, Spark
+- 2026-10-26T01:43:39 · training · jungfrau@mx-spark-01 — train_run.py
+- 2026-10-26T09:16:24 · measured · jungfrau@mx-spark-01
+- 2026-10-26T09:16:24 · gate_failed · jungfrau@mx-spark-01 — {"baseline": "/home/jungfrau/AareLC_train/models/registry/yolo26s-seg-v5-s-noflip-mask1-2026-10-25/gold_report.json", "f1_delta": -0.0448, "mask_iou_delta": -0.0327, "mask_score": 0.6503, "mask_score_baseline": 0.6831, "matched_only_iou_delta": 0.0043, "class_f1_delta": {"loop_all": -0.0483, "pin": -0.0221}, "pass_point_rule": false, "rule": "paired bootstrap", "bootstrap": {"frames": 335, "mounts": 300, "draws": 2000, "f1": {"delta": -0.0448, "lo": -0.074, "hi": -0.0181, "win": 0.001}, "mask_score": {"delta": -0.0327, "lo": -0.0544, "hi": -0.0126, "win": 0.001}}, "pass": false}