2026-06-03 16:31:49.243860: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2026-06-03 16:31:52.095285: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT 2026-06-03 16:31:59.966324: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1639] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 46732 MB memory: -> device: 0, name: NVIDIA RTX A6000, pci bus id: 0000:41:00.0, compute capability: 8.6 WARNING:absl:Skipping variable loading for optimizer 'RMSprop', because it has 1 variables whereas the saved optimizer has 301 variables. 2026-06-03 16:32:05.682868: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:432] Loaded cuDNN version 8907 2026-06-03 16:32:05.816842: I tensorflow/tsl/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory 2026-06-03 16:32:06.284320: I tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:606] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once. Building model… Weights loaded. Loading 92 test images… Predicting… 1/12 [=>............................] - ETA: 21s 6/12 [==============>...............] - ETA: 0s  11/12 [==========================>...] - ETA: 0s 12/12 [==============================] - ETA: 0s 12/12 [==============================] - 3s 70ms/step ============================================================ ROCKGLACIERNET — EVALUATION SUMMARY (baseline 128x128) ============================================================ Test images : 92 Best threshold by IoU : 0.65 GLOBAL (pixel-pooled) metrics @ best threshold: precision : 0.6935 recall : 0.8028 f1 : 0.7442 iou : 0.5926 dice : 0.7442 PER-IMAGE IoU (@ t=0.5): mean : 0.6622 median : 0.6957 Threshold sweep: thresh precision recall f1 iou dice ------------------------------------------------------------ 0.3 0.6169 0.8922 0.7294 0.5741 0.7294 0.35 0.6276 0.8812 0.7331 0.5786 0.7331 0.4 0.6382 0.8699 0.7363 0.5826 0.7363 0.45 0.6483 0.8583 0.7387 0.5856 0.7387 0.5 0.6582 0.8463 0.7405 0.5879 0.7405 0.55 0.6689 0.8332 0.7421 0.5899 0.7421 0.6 0.6807 0.8193 0.7436 0.5919 0.7436 0.65 0.6935 0.8028 0.7442 0.5926 0.7442 <- best IoU 0.7 0.7081 0.7822 0.7433 0.5915 0.7433 10 WORST images by IoU (@ t=0.5): image_id orig_H orig_W glacier_frac_gt iou precision recall 0242 130 120 0.1247 0.2662 0.2758 0.8841 0286 289 177 0.3099 0.3416 0.3696 0.8186 0450 80 82 0.0709 0.3615 0.3849 0.8559 0288 250 115 0.2682 0.3783 0.5689 0.5304 0301 125 84 0.1251 0.3966 0.4103 0.9224 0491 141 74 0.1455 0.4199 0.4927 0.7395 0615 91 60 0.0974 0.4655 0.5520 0.7481 0234 98 110 0.2587 0.4680 0.8010 0.5296 0032 55 60 0.0436 0.4833 0.5472 0.8056 0268 146 135 0.2199 0.4892 0.6085 0.7138 Output dir: outputs/predictions/baseline_model_output Files: per_image_metrics.csv, evaluation_metrics.csv, confusion_matrix.csv Saved 20 comparison PNGs (10 worst + 10 best). Evaluation complete.