diff --git a/.gitignore b/.gitignore
index 16d88fce..aa10c52c 100644
--- a/.gitignore
+++ b/.gitignore
@@ -3,4 +3,5 @@ data/
 *.zip
 tiles/
 tiles_annotated/
-tiles_lanov/
\ No newline at end of file
+tiles_lanov/
+dataset_lanov.zip
\ No newline at end of file
diff --git a/02_training.ipynb b/02_training.ipynb
index f816c0c6..d7f7eac5 100644
--- a/02_training.ipynb
+++ b/02_training.ipynb
@@ -76,368 +76,10 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": null,
    "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Trénink na: cuda\n",
-      "Ultralytics 8.4.40 🚀 Python-3.11.10 torch-2.4.1+cu124 CUDA:0 (NVIDIA GeForce RTX 4090, 24210MiB)\n",
-      "\u001b[34m\u001b[1mengine/trainer: \u001b[0magnostic_nms=False, amp=True, angle=1.0, augment=False, auto_augment=randaugment, batch=16, bgr=0.0, box=7.5, cache=False, cfg=None, classes=None, close_mosaic=10, cls=0.5, cls_pw=0.0, compile=False, conf=None, copy_paste=0.0, copy_paste_mode=flip, cos_lr=False, cutmix=0.0, data=data/yolo_visdrone/dataset.yaml, degrees=15.0, deterministic=True, device=0, dfl=1.5, dnn=False, dropout=0.0, dynamic=False, embed=None, end2end=None, epochs=50, erasing=0.4, exist_ok=False, fliplr=0.5, flipud=0.5, format=torchscript, fraction=1.0, freeze=None, half=False, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, imgsz=640, int8=False, iou=0.7, keras=False, kobj=1.0, line_width=None, lr0=0.01, lrf=0.01, mask_ratio=4, max_det=300, mixup=0.0, mode=train, model=yolov8s.pt, momentum=0.937, mosaic=1.0, multi_scale=0.0, name=visdrone_vehicles, nbs=64, nms=False, opset=None, optimize=False, optimizer=auto, overlap_mask=True, patience=10, perspective=0.0, plots=True, pose=12.0, pretrained=True, profile=False, project=runs/train, rect=False, resume=False, retina_masks=False, rle=1.0, save=True, save_conf=False, save_crop=False, save_dir=/workspace/runs/detect/runs/train/visdrone_vehicles, save_frames=False, save_json=False, save_period=10, save_txt=False, scale=0.5, seed=0, shear=0.0, show=False, show_boxes=True, show_conf=True, show_labels=True, simplify=True, single_cls=False, source=None, split=val, stream_buffer=False, task=detect, time=None, tracker=botsort.yaml, translate=0.1, val=True, verbose=True, vid_stride=1, visualize=False, warmup_bias_lr=0.1, warmup_epochs=3.0, warmup_momentum=0.8, weight_decay=0.0005, workers=8, workspace=None\n",
-      "Downloading https://ultralytics.com/assets/Arial.ttf to '/root/.config/Ultralytics/Arial.ttf': 100% ━━━━━━━━━━━━ 755.1KB 61.9MB/s 0.0s\n",
-      "Overriding model.yaml nc=80 with nc=4\n",
-      "\n",
-      "                   from  n    params  module                                       arguments                     \n",
-      "  0                  -1  1       928  ultralytics.nn.modules.conv.Conv             [3, 32, 3, 2]                 \n",
-      "  1                  -1  1     18560  ultralytics.nn.modules.conv.Conv             [32, 64, 3, 2]                \n",
-      "  2                  -1  1     29056  ultralytics.nn.modules.block.C2f             [64, 64, 1, True]             \n",
-      "  3                  -1  1     73984  ultralytics.nn.modules.conv.Conv             [64, 128, 3, 2]               \n",
-      "  4                  -1  2    197632  ultralytics.nn.modules.block.C2f             [128, 128, 2, True]           \n",
-      "  5                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]              \n",
-      "  6                  -1  2    788480  ultralytics.nn.modules.block.C2f             [256, 256, 2, True]           \n",
-      "  7                  -1  1   1180672  ultralytics.nn.modules.conv.Conv             [256, 512, 3, 2]              \n",
-      "  8                  -1  1   1838080  ultralytics.nn.modules.block.C2f             [512, 512, 1, True]           \n",
-      "  9                  -1  1    656896  ultralytics.nn.modules.block.SPPF            [512, 512, 5]                 \n",
-      " 10                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          \n",
-      " 11             [-1, 6]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           \n",
-      " 12                  -1  1    591360  ultralytics.nn.modules.block.C2f             [768, 256, 1]                 \n",
-      " 13                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          \n",
-      " 14             [-1, 4]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           \n",
-      " 15                  -1  1    148224  ultralytics.nn.modules.block.C2f             [384, 128, 1]                 \n",
-      " 16                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]              \n",
-      " 17            [-1, 12]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           \n",
-      " 18                  -1  1    493056  ultralytics.nn.modules.block.C2f             [384, 256, 1]                 \n",
-      " 19                  -1  1    590336  ultralytics.nn.modules.conv.Conv             [256, 256, 3, 2]              \n",
-      " 20             [-1, 9]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           \n",
-      " 21                  -1  1   1969152  ultralytics.nn.modules.block.C2f             [768, 512, 1]                 \n",
-      " 22        [15, 18, 21]  1   2117596  ultralytics.nn.modules.head.Detect           [4, 16, None, [128, 256, 512]]\n",
-      "Model summary: 130 layers, 11,137,148 parameters, 11,137,132 gradients, 28.7 GFLOPs\n",
-      "\n",
-      "Transferred 349/355 items from pretrained weights\n",
-      "Freezing layer 'model.22.dfl.conv.weight'\n",
-      "\u001b[34m\u001b[1mAMP: \u001b[0mrunning Automatic Mixed Precision (AMP) checks...\n",
-      "Downloading https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26n.pt to 'yolo26n.pt': 100% ━━━━━━━━━━━━ 5.3MB 46.4MB/s 0.1s\n",
-      "\u001b[34m\u001b[1mAMP: \u001b[0mchecks passed ✅\n",
-      "\u001b[34m\u001b[1mtrain: \u001b[0mFast image access ✅ (ping: 0.6±0.2 ms, read: 48.3±24.4 MB/s, size: 260.7 KB)\n",
-      "\u001b[34m\u001b[1mtrain: \u001b[0mScanning /workspace/data/yolo_visdrone/train/labels... 6471 images, 279 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 6471/6471 514.6it/s 12.6s\n",
-      "\u001b[34m\u001b[1mtrain: \u001b[0m/workspace/data/yolo_visdrone/train/images/0000140_00118_d_0000002.jpg: 1 duplicate labels removed\n",
-      "\u001b[34m\u001b[1mtrain: \u001b[0mNew cache created: /workspace/data/yolo_visdrone/train/labels.cache\n",
-      "\u001b[34m\u001b[1mval: \u001b[0mFast image access ✅ (ping: 0.4±0.4 ms, read: 48.4±15.9 MB/s, size: 142.1 KB)\n",
-      "\u001b[34m\u001b[1mval: \u001b[0mScanning /workspace/data/yolo_visdrone/val/labels... 548 images, 29 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 548/548 365.1it/s 1.5s\n",
-      "\u001b[34m\u001b[1mval: \u001b[0mNew cache created: /workspace/data/yolo_visdrone/val/labels.cache\n",
-      "\u001b[34m\u001b[1moptimizer:\u001b[0m 'optimizer=auto' found, ignoring 'lr0=0.01' and 'momentum=0.937' and determining best 'optimizer', 'lr0' and 'momentum' automatically... \n",
-      "\u001b[34m\u001b[1moptimizer:\u001b[0m AdamW(lr=0.00125, momentum=0.9) with parameter groups 57 weight(decay=0.0), 64 weight(decay=0.0005), 63 bias(decay=0.0)\n",
-      "Plotting labels to /workspace/runs/detect/runs/train/visdrone_vehicles/labels.jpg... \n",
-      "Image sizes 640 train, 640 val\n",
-      "Using 8 dataloader workers\n",
-      "Logging results to \u001b[1m/workspace/runs/detect/runs/train/visdrone_vehicles\u001b[0m\n",
-      "Starting training for 50 epochs...\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       1/50      6.74G      1.681      1.528      1.078        120        640: 100% ━━━━━━━━━━━━ 405/405 11.2it/s 36.0s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 4.4it/s 4.1s\n",
-      "                   all        548      17040      0.364      0.302      0.271      0.148\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       2/50      7.48G      1.561      1.225      1.018        411        640: 100% ━━━━━━━━━━━━ 405/405 13.3it/s 30.3s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 5.3it/s 3.4s\n",
-      "                   all        548      17040      0.479      0.293      0.301      0.178\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       3/50       7.5G      1.538      1.186      1.007        316        640: 100% ━━━━━━━━━━━━ 405/405 13.5it/s 30.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 4.7it/s 3.8s\n",
-      "                   all        548      17040      0.384      0.342      0.304      0.167\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       4/50       7.5G      1.503      1.155      0.998        229        640: 100% ━━━━━━━━━━━━ 405/405 13.1it/s 31.0s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 5.0it/s 3.6s\n",
-      "                   all        548      17040      0.487       0.33      0.329      0.206\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       5/50       7.5G      1.465      1.105     0.9862        296        640: 100% ━━━━━━━━━━━━ 405/405 13.5it/s 30.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 5.4it/s 3.3s\n",
-      "                   all        548      17040      0.489      0.348      0.345      0.198\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       6/50       7.5G      1.453      1.075     0.9789        307        640: 100% ━━━━━━━━━━━━ 405/405 13.6it/s 29.9s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 4.9it/s 3.7s\n",
-      "                   all        548      17040       0.55      0.362      0.385       0.21\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       7/50       7.5G      1.433      1.058     0.9745        352        640: 100% ━━━━━━━━━━━━ 405/405 12.6it/s 32.2s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 5.3it/s 3.4s\n",
-      "                   all        548      17040      0.526      0.367      0.385      0.238\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       8/50       7.5G      1.411      1.033     0.9689        130        640: 100% ━━━━━━━━━━━━ 405/405 12.7it/s 31.9s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.9it/s 2.6s\n",
-      "                   all        548      17040      0.544       0.38      0.396      0.251\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       9/50       7.5G      1.391      1.016     0.9621        362        640: 100% ━━━━━━━━━━━━ 405/405 12.2it/s 33.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.9it/s 2.6s\n",
-      "                   all        548      17040      0.533      0.391      0.408      0.257\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      10/50       7.5G       1.37          1     0.9538        251        640: 100% ━━━━━━━━━━━━ 405/405 12.1it/s 33.5s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.2it/s 2.5s\n",
-      "                   all        548      17040       0.56      0.366      0.383      0.239\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      11/50       7.5G      1.359     0.9788     0.9528        315        640: 100% ━━━━━━━━━━━━ 405/405 13.0it/s 31.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 5.9it/s 3.0s\n",
-      "                   all        548      17040      0.528      0.393      0.408      0.248\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      12/50      8.27G      1.349     0.9773     0.9515        194        640: 100% ━━━━━━━━━━━━ 405/405 12.7it/s 32.0s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.8it/s 2.3s\n",
-      "                   all        548      17040      0.569      0.408      0.424      0.266\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      13/50      8.27G      1.338     0.9568     0.9487        302        640: 100% ━━━━━━━━━━━━ 405/405 12.9it/s 31.4s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.4it/s 2.8s\n",
-      "                   all        548      17040      0.586      0.403       0.43      0.272\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      14/50      8.27G      1.329     0.9541     0.9422        212        640: 100% ━━━━━━━━━━━━ 405/405 12.1it/s 33.4s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.1it/s 3.0s\n",
-      "                   all        548      17040      0.583      0.421      0.452      0.282\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      15/50      8.27G      1.327     0.9459     0.9386        312        640: 100% ━━━━━━━━━━━━ 405/405 13.0it/s 31.2s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.9it/s 2.6s\n",
-      "                   all        548      17040      0.592      0.424      0.453      0.289\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      16/50      8.27G      1.316     0.9352     0.9406        391        640: 100% ━━━━━━━━━━━━ 405/405 13.0it/s 31.0s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.4it/s 2.8s\n",
-      "                   all        548      17040      0.577      0.417      0.442      0.276\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      17/50      8.27G      1.302     0.9241     0.9354        289        640: 100% ━━━━━━━━━━━━ 405/405 13.0it/s 31.2s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.8it/s 2.7s\n",
-      "                   all        548      17040      0.615       0.41      0.455      0.286\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      18/50      8.27G      1.296     0.9194     0.9328        343        640: 100% ━━━━━━━━━━━━ 405/405 12.1it/s 33.4s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.3it/s 2.8s\n",
-      "                   all        548      17040      0.613      0.395      0.436       0.28\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      19/50      8.27G      1.285     0.9064     0.9314        176        640: 100% ━━━━━━━━━━━━ 405/405 12.5it/s 32.3s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 5.9it/s 3.1s\n",
-      "                   all        548      17040      0.596      0.427      0.454      0.287\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      20/50      8.27G      1.272     0.8993     0.9273        419        640: 100% ━━━━━━━━━━━━ 405/405 12.5it/s 32.4s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.8it/s 2.6s\n",
-      "                   all        548      17040      0.603      0.414      0.452      0.294\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      21/50      8.27G      1.266     0.8915     0.9256        368        640: 100% ━━━━━━━━━━━━ 405/405 12.4it/s 32.7s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.2it/s 2.9s\n",
-      "                   all        548      17040      0.635      0.433      0.479      0.303\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      22/50      8.27G      1.252     0.8861     0.9235        399        640: 100% ━━━━━━━━━━━━ 405/405 12.3it/s 32.8s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.2it/s 2.5s\n",
-      "                   all        548      17040      0.639      0.433      0.472      0.299\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      23/50      8.27G      1.249     0.8802     0.9229        323        640: 100% ━━━━━━━━━━━━ 405/405 12.7it/s 31.9s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.1it/s 2.5s\n",
-      "                   all        548      17040      0.616      0.427      0.467      0.305\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      24/50      8.27G      1.232     0.8676     0.9185        515        640: 100% ━━━━━━━━━━━━ 405/405 14.1it/s 28.7s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.3it/s 2.9s\n",
-      "                   all        548      17040       0.64       0.42      0.471      0.308\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      25/50      8.27G       1.23     0.8628      0.917        293        640: 100% ━━━━━━━━━━━━ 405/405 13.8it/s 29.4s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.9it/s 2.6s\n",
-      "                   all        548      17040      0.592       0.45      0.478       0.31\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      26/50       9.3G      1.231     0.8637     0.9194        339        640: 100% ━━━━━━━━━━━━ 405/405 12.1it/s 33.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.7it/s 2.7s\n",
-      "                   all        548      17040      0.635      0.418      0.467      0.305\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      27/50       9.3G      1.217     0.8563     0.9122        342        640: 100% ━━━━━━━━━━━━ 405/405 12.9it/s 31.4s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.5it/s 2.8s\n",
-      "                   all        548      17040      0.643      0.436      0.484      0.314\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      28/50       9.3G      1.213     0.8475     0.9091        387        640: 100% ━━━━━━━━━━━━ 405/405 13.5it/s 30.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.2it/s 2.9s\n",
-      "                   all        548      17040      0.622      0.449      0.486       0.31\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      29/50       9.3G       1.21     0.8415     0.9099        393        640: 100% ━━━━━━━━━━━━ 405/405 13.5it/s 30.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.3it/s 2.8s\n",
-      "                   all        548      17040      0.604      0.452      0.488       0.32\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      30/50       9.3G      1.203     0.8385     0.9082        296        640: 100% ━━━━━━━━━━━━ 405/405 13.9it/s 29.2s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.8it/s 2.6s\n",
-      "                   all        548      17040      0.604      0.452      0.487      0.311\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      31/50       9.3G      1.186     0.8276     0.9057        187        640: 100% ━━━━━━━━━━━━ 405/405 12.9it/s 31.5s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.6it/s 2.4s\n",
-      "                   all        548      17040      0.607       0.46      0.495      0.322\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      32/50       9.3G      1.194     0.8245     0.9052        312        640: 100% ━━━━━━━━━━━━ 405/405 12.7it/s 32.0s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.9it/s 2.6s\n",
-      "                   all        548      17040      0.652      0.451      0.502      0.327\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      33/50       9.3G      1.185     0.8258     0.9068        385        640: 100% ━━━━━━━━━━━━ 405/405 13.3it/s 30.4s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.5it/s 2.8s\n",
-      "                   all        548      17040      0.603      0.474      0.499      0.322\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      34/50       9.3G      1.181     0.8198     0.9028        300        640: 100% ━━━━━━━━━━━━ 405/405 13.2it/s 30.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.1it/s 2.5s\n",
-      "                   all        548      17040      0.651      0.451        0.5      0.326\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      35/50       9.3G      1.175     0.8094     0.9012        312        640: 100% ━━━━━━━━━━━━ 405/405 13.1it/s 30.9s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 8.8it/s 2.0s\n",
-      "                   all        548      17040      0.622      0.455      0.494      0.326\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      36/50       9.3G      1.167     0.8011     0.9009        399        640: 100% ━━━━━━━━━━━━ 405/405 14.2it/s 28.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.5it/s 2.4s\n",
-      "                   all        548      17040      0.634      0.452      0.499       0.33\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      37/50       9.3G      1.161     0.7976     0.9003        318        640: 100% ━━━━━━━━━━━━ 405/405 13.5it/s 29.9s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.4it/s 2.8s\n",
-      "                   all        548      17040      0.625      0.475      0.509      0.335\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      38/50       9.3G      1.164     0.7973     0.8987        280        640: 100% ━━━━━━━━━━━━ 405/405 13.5it/s 30.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.7it/s 2.3s\n",
-      "                   all        548      17040      0.625      0.482      0.517       0.34\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      39/50       9.3G      1.155     0.7871     0.8957        475        640: 100% ━━━━━━━━━━━━ 405/405 13.9it/s 29.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 9.0it/s 2.0s\n",
-      "                   all        548      17040       0.64      0.466      0.512      0.334\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      40/50       9.3G      1.151     0.7872     0.8961        274        640: 100% ━━━━━━━━━━━━ 405/405 14.1it/s 28.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 9.6it/s 1.9s\n",
-      "                   all        548      17040      0.644      0.471       0.51      0.334\n",
-      "Closing dataloader mosaic\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      41/50       9.3G      1.115     0.7576     0.8965        134        640: 100% ━━━━━━━━━━━━ 405/405 13.3it/s 30.5s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.6it/s 2.4s\n",
-      "                   all        548      17040      0.645      0.466      0.505       0.33\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      42/50       9.3G      1.105     0.7458      0.895        193        640: 100% ━━━━━━━━━━━━ 405/405 13.4it/s 30.1s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 8.4it/s 2.1s\n",
-      "                   all        548      17040      0.659      0.451       0.51      0.334\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      43/50       9.3G      1.099     0.7388     0.8945        119        640: 100% ━━━━━━━━━━━━ 405/405 15.1it/s 26.8s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 9.1it/s 2.0s\n",
-      "                   all        548      17040      0.589      0.486      0.505      0.335\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      44/50       9.3G      1.091     0.7312     0.8914        272        640: 100% ━━━━━━━━━━━━ 405/405 15.0it/s 27.0s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 8.9it/s 2.0s\n",
-      "                   all        548      17040      0.619      0.476      0.504      0.335\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      45/50       9.3G      1.086     0.7252     0.8911        180        640: 100% ━━━━━━━━━━━━ 405/405 14.5it/s 27.9s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 8.7it/s 2.1s\n",
-      "                   all        548      17040      0.638      0.468      0.506      0.338\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      46/50       9.3G      1.084     0.7237     0.8881        140        640: 100% ━━━━━━━━━━━━ 405/405 14.9it/s 27.2s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 8.4it/s 2.1s\n",
-      "                   all        548      17040      0.669      0.453       0.51      0.336\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      47/50       9.3G      1.078     0.7178     0.8877        187        640: 100% ━━━━━━━━━━━━ 405/405 14.2it/s 28.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 8.0it/s 2.3s\n",
-      "                   all        548      17040      0.663      0.458      0.512      0.339\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      48/50       9.3G      1.072     0.7115     0.8862        265        640: 100% ━━━━━━━━━━━━ 405/405 13.4it/s 30.2s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 7.9it/s 2.3s\n",
-      "                   all        548      17040      0.646      0.465      0.512      0.341\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      49/50       9.3G      1.073     0.7076     0.8848        238        640: 100% ━━━━━━━━━━━━ 405/405 13.8it/s 29.3s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 8.4it/s 2.1s\n",
-      "                   all        548      17040      0.632      0.468      0.515      0.343\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      50/50       9.3G      1.062     0.7007     0.8852        221        640: 100% ━━━━━━━━━━━━ 405/405 14.6it/s 27.7s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 9.5it/s 1.9s\n",
-      "                   all        548      17040      0.649      0.463      0.514      0.343\n",
-      "\n",
-      "50 epochs completed in 0.477 hours.\n",
-      "Optimizer stripped from /workspace/runs/detect/runs/train/visdrone_vehicles/weights/last.pt, 22.5MB\n",
-      "Optimizer stripped from /workspace/runs/detect/runs/train/visdrone_vehicles/weights/best.pt, 22.5MB\n",
-      "\n",
-      "Validating /workspace/runs/detect/runs/train/visdrone_vehicles/weights/best.pt...\n",
-      "Ultralytics 8.4.40 🚀 Python-3.11.10 torch-2.4.1+cu124 CUDA:0 (NVIDIA GeForce RTX 4090, 24210MiB)\n",
-      "Model summary (fused): 73 layers, 11,127,132 parameters, 0 gradients, 28.4 GFLOPs\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 18/18 6.3it/s 2.9s\n",
-      "                   all        548      17040      0.648      0.463      0.514      0.343\n",
-      "                   car        515      14064      0.752      0.751      0.772       0.51\n",
-      "                   van        421       1975        0.6      0.368      0.425      0.285\n",
-      "                 truck        266        750      0.579        0.3      0.342      0.223\n",
-      "                   bus        131        251      0.661      0.435      0.515      0.353\n",
-      "Speed: 0.1ms preprocess, 0.4ms inference, 0.0ms loss, 2.4ms postprocess per image\n",
-      "Results saved to \u001b[1m/workspace/runs/detect/runs/train/visdrone_vehicles\u001b[0m\n"
-     ]
-    }
-   ],
-   "source": [
-    "import torch\n",
-    "\n",
-    "# Detekce dostupného zařízení\n",
-    "if torch.cuda.is_available():\n",
-    "    DEVICE = \"cuda\"\n",
-    "elif torch.backends.mps.is_available():\n",
-    "    DEVICE = \"mps\"\n",
-    "else:\n",
-    "    DEVICE = \"cpu\"\n",
-    "\n",
-    "print(f\"Trénink na: {DEVICE}\")\n",
-    "\n",
-    "results = model.train(\n",
-    "    data=str(YAML),\n",
-    "    epochs=50,\n",
-    "    imgsz=640,\n",
-    "    batch=16,\n",
-    "    device=DEVICE,\n",
-    "    name=\"visdrone_vehicles\",\n",
-    "    project=\"runs/train\",\n",
-    "    patience=10,           # early stopping\n",
-    "    save=True,\n",
-    "    save_period=10,\n",
-    "    val=True,\n",
-    "    plots=True,\n",
-    "    # Augmentace pro letecké snímky\n",
-    "    degrees=15.0,          # rotace\n",
-    "    flipud=0.5,            # vertikální flip\n",
-    "    fliplr=0.5,\n",
-    "    mosaic=1.0,\n",
-    "    scale=0.5,\n",
-    ")"
-   ]
+   "outputs": [],
+   "source": "import torch\nfrom pathlib import Path\n\n# Detekce dostupného zařízení\nif torch.cuda.is_available():\n    DEVICE = \"cuda\"\nelif torch.backends.mps.is_available():\n    DEVICE = \"mps\"\nelse:\n    DEVICE = \"cpu\"\n\nprint(f\"Trénink na: {DEVICE}\")\n\nresults = model.train(\n    data=str(YAML),\n    epochs=50,\n    imgsz=640,\n    batch=16,\n    device=DEVICE,\n    name=\"visdrone_vehicles\",\n    project=str(Path(\"runs/train\").resolve()),\n    patience=10,           # early stopping\n    save=True,\n    save_period=10,\n    val=True,\n    plots=True,\n    # Augmentace pro letecké snímky\n    degrees=15.0,          # rotace\n    flipud=0.5,            # vertikální flip\n    fliplr=0.5,\n    mosaic=1.0,\n    scale=0.5,\n)"
   },
   {
    "cell_type": "code",
@@ -475,241 +117,17 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": null,
    "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Ultralytics 8.4.40 🚀 Python-3.11.10 torch-2.4.1+cu124 CUDA:0 (NVIDIA GeForce RTX 4090, 24210MiB)\n",
-      "Model summary (fused): 73 layers, 11,127,132 parameters, 0 gradients, 28.4 GFLOPs\n",
-      "\u001b[34m\u001b[1mval: \u001b[0mFast image access ✅ (ping: 0.3±0.2 ms, read: 97.7±33.9 MB/s, size: 126.7 KB)\n",
-      "\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 35/35 11.2it/s 3.1s\n",
-      "                   all        548      17040      0.651      0.465      0.515      0.344\n",
-      "                   car        515      14064      0.753      0.751      0.774      0.511\n",
-      "                   van        421       1975      0.604      0.369      0.426      0.286\n",
-      "                 truck        266        750      0.578        0.3      0.342      0.224\n",
-      "                   bus        131        251      0.667       0.44      0.518      0.354\n",
-      "Speed: 0.6ms preprocess, 1.3ms inference, 0.0ms loss, 0.6ms postprocess per image\n",
-      "Results saved to \u001b[1m/workspace/runs/detect/val\u001b[0m\n",
-      "\n",
-      "mAP50:    0.5151\n",
-      "mAP50-95: 0.3438\n",
-      "  AP50[car]: 0.7742\n",
-      "  AP50[van]: 0.4262\n",
-      "  AP50[truck]: 0.3418\n",
-      "  AP50[bus]: 0.5181\n"
-     ]
-    }
-   ],
-   "source": [
-    "# Validace nejlepšího modelu\n",
-    "best_weights = Path(\"runs/detect/runs/train/visdrone_vehicles/weights/best.pt\")\n",
-    "assert best_weights.exists(), \"Trénink ještě neproběhl nebo selhal\"\n",
-    "\n",
-    "model_best = YOLO(str(best_weights))\n",
-    "val_results = model_best.val(data=str(YAML), imgsz=640, split=\"val\")\n",
-    "\n",
-    "print(f\"\\nmAP50:    {val_results.box.map50:.4f}\")\n",
-    "print(f\"mAP50-95: {val_results.box.map:.4f}\")\n",
-    "for i, cls in enumerate([\"car\", \"van\", \"truck\", \"bus\"]):\n",
-    "    ap = val_results.box.ap50[i] if i < len(val_results.box.ap50) else float('nan')\n",
-    "    print(f\"  AP50[{cls}]: {ap:.4f}\")"
-   ]
+   "outputs": [],
+   "source": "# Validace nejlepšího modelu\nbest_weights = Path(\"runs/train/visdrone_vehicles/weights/best.pt\")\nassert best_weights.exists(), \"Trénink ještě neproběhl nebo selhal\"\n\nmodel_best = YOLO(str(best_weights))\nval_results = model_best.val(data=str(YAML), imgsz=640, split=\"val\")\n\nprint(f\"\\nmAP50:    {val_results.box.map50:.4f}\")\nprint(f\"mAP50-95: {val_results.box.map:.4f}\")\nfor i, cls in enumerate([\"car\", \"van\", \"truck\", \"bus\"]):\n    ap = val_results.box.ap50[i] if i < len(val_results.box.ap50) else float('nan')\n    print(f\"  AP50[{cls}]: {ap:.4f}\")"
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": null,
    "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Fine-tuning na: lanov/dataset.yaml\n",
-      "Ultralytics 8.4.40 🚀 Python-3.11.10 torch-2.4.1+cu124 CUDA:0 (NVIDIA GeForce RTX 4090, 24210MiB)\n",
-      "\u001b[34m\u001b[1mengine/trainer: \u001b[0magnostic_nms=False, amp=True, angle=1.0, augment=False, auto_augment=randaugment, batch=32, bgr=0.0, box=7.5, cache=False, cfg=None, classes=None, close_mosaic=10, cls=0.5, cls_pw=0.0, compile=False, conf=None, copy_paste=0.0, copy_paste_mode=flip, cos_lr=False, cutmix=0.0, data=lanov/dataset.yaml, degrees=10.0, deterministic=True, device=0, dfl=1.5, dnn=False, dropout=0.0, dynamic=False, embed=None, end2end=None, epochs=30, erasing=0.4, exist_ok=False, fliplr=0.5, flipud=0.5, format=torchscript, fraction=1.0, freeze=None, half=False, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, imgsz=256, int8=False, iou=0.7, keras=False, kobj=1.0, line_width=None, lr0=0.001, lrf=0.01, mask_ratio=4, max_det=300, mixup=0.0, mode=train, model=runs/detect/runs/train/visdrone_vehicles/weights/best.pt, momentum=0.937, mosaic=0.5, multi_scale=0.0, name=lanov_finetune-2, nbs=64, nms=False, opset=None, optimize=False, optimizer=auto, overlap_mask=True, patience=10, perspective=0.0, plots=True, pose=12.0, pretrained=True, profile=False, project=runs/detect/runs/train, rect=False, resume=False, retina_masks=False, rle=1.0, save=True, save_conf=False, save_crop=False, save_dir=/workspace/runs/detect/runs/detect/runs/train/lanov_finetune-2, save_frames=False, save_json=False, save_period=-1, save_txt=False, scale=0.5, seed=0, shear=0.0, show=False, show_boxes=True, show_conf=True, show_labels=True, simplify=True, single_cls=False, source=None, split=val, stream_buffer=False, task=detect, time=None, tracker=botsort.yaml, translate=0.1, val=True, verbose=True, vid_stride=1, visualize=False, warmup_bias_lr=0.1, warmup_epochs=3.0, warmup_momentum=0.8, weight_decay=0.0005, workers=8, workspace=None\n",
-      "\n",
-      "                   from  n    params  module                                       arguments                     \n",
-      "  0                  -1  1       928  ultralytics.nn.modules.conv.Conv             [3, 32, 3, 2]                 \n",
-      "  1                  -1  1     18560  ultralytics.nn.modules.conv.Conv             [32, 64, 3, 2]                \n",
-      "  2                  -1  1     29056  ultralytics.nn.modules.block.C2f             [64, 64, 1, True]             \n",
-      "  3                  -1  1     73984  ultralytics.nn.modules.conv.Conv             [64, 128, 3, 2]               \n",
-      "  4                  -1  2    197632  ultralytics.nn.modules.block.C2f             [128, 128, 2, True]           \n",
-      "  5                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]              \n",
-      "  6                  -1  2    788480  ultralytics.nn.modules.block.C2f             [256, 256, 2, True]           \n",
-      "  7                  -1  1   1180672  ultralytics.nn.modules.conv.Conv             [256, 512, 3, 2]              \n",
-      "  8                  -1  1   1838080  ultralytics.nn.modules.block.C2f             [512, 512, 1, True]           \n",
-      "  9                  -1  1    656896  ultralytics.nn.modules.block.SPPF            [512, 512, 5]                 \n",
-      " 10                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          \n",
-      " 11             [-1, 6]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           \n",
-      " 12                  -1  1    591360  ultralytics.nn.modules.block.C2f             [768, 256, 1]                 \n",
-      " 13                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          \n",
-      " 14             [-1, 4]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           \n",
-      " 15                  -1  1    148224  ultralytics.nn.modules.block.C2f             [384, 128, 1]                 \n",
-      " 16                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]              \n",
-      " 17            [-1, 12]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           \n",
-      " 18                  -1  1    493056  ultralytics.nn.modules.block.C2f             [384, 256, 1]                 \n",
-      " 19                  -1  1    590336  ultralytics.nn.modules.conv.Conv             [256, 256, 3, 2]              \n",
-      " 20             [-1, 9]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           \n",
-      " 21                  -1  1   1969152  ultralytics.nn.modules.block.C2f             [768, 512, 1]                 \n",
-      " 22        [15, 18, 21]  1   2117596  ultralytics.nn.modules.head.Detect           [4, 16, None, [128, 256, 512]]\n",
-      "Model summary: 130 layers, 11,137,148 parameters, 11,137,132 gradients, 28.7 GFLOPs\n",
-      "\n",
-      "Transferred 355/355 items from pretrained weights\n",
-      "Freezing layer 'model.22.dfl.conv.weight'\n",
-      "\u001b[34m\u001b[1mAMP: \u001b[0mrunning Automatic Mixed Precision (AMP) checks...\n",
-      "\u001b[34m\u001b[1mAMP: \u001b[0mchecks passed ✅\n",
-      "\u001b[34m\u001b[1mtrain: \u001b[0mFast image access ✅ (ping: 0.3±0.1 ms, read: 7.1±1.5 MB/s, size: 19.3 KB)\n",
-      "\u001b[34m\u001b[1mtrain: \u001b[0mScanning /workspace/lanov/train/labels... 253 images, 50 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 253/253 750.0it/s 0.3s\n",
-      "\u001b[34m\u001b[1mtrain: \u001b[0mNew cache created: /workspace/lanov/train/labels.cache\n",
-      "\u001b[34m\u001b[1mval: \u001b[0mFast image access ✅ (ping: 0.2±0.0 ms, read: 4.8±2.1 MB/s, size: 14.6 KB)\n",
-      "\u001b[34m\u001b[1mval: \u001b[0mScanning /workspace/lanov/val/labels... 100 images, 50 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 100/100 470.1it/s 0.2s\n",
-      "\u001b[34m\u001b[1mval: \u001b[0mNew cache created: /workspace/lanov/val/labels.cache\n",
-      "\u001b[34m\u001b[1moptimizer:\u001b[0m 'optimizer=auto' found, ignoring 'lr0=0.001' and 'momentum=0.937' and determining best 'optimizer', 'lr0' and 'momentum' automatically... \n",
-      "\u001b[34m\u001b[1moptimizer:\u001b[0m AdamW(lr=0.00125, momentum=0.9) with parameter groups 57 weight(decay=0.0), 64 weight(decay=0.0005), 63 bias(decay=0.0)\n",
-      "Plotting labels to /workspace/runs/detect/runs/detect/runs/train/lanov_finetune-2/labels.jpg... \n",
-      "Image sizes 256 train, 256 val\n",
-      "Using 8 dataloader workers\n",
-      "Logging results to \u001b[1m/workspace/runs/detect/runs/detect/runs/train/lanov_finetune-2\u001b[0m\n",
-      "Starting training for 30 epochs...\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       1/30      1.46G       3.22      3.365      1.341         78        256: 100% ━━━━━━━━━━━━ 8/8 3.5it/s 2.3s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 6.8it/s 0.3s\n",
-      "                   all        100        271      0.248      0.142      0.114      0.035\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       2/30       1.5G      3.126      2.779      1.245        111        256: 100% ━━━━━━━━━━━━ 8/8 13.7it/s 0.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 14.2it/s 0.1s\n",
-      "                   all        100        271       0.17      0.261      0.165      0.054\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       3/30       1.5G      2.761      1.844       1.18        106        256: 100% ━━━━━━━━━━━━ 8/8 11.9it/s 0.7s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 11.8it/s 0.2s\n",
-      "                   all        100        271       0.47      0.289      0.187     0.0661\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       4/30       1.5G      2.577      1.633      1.037        167        256: 100% ━━━━━━━━━━━━ 8/8 11.6it/s 0.7s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 12.4it/s 0.2s\n",
-      "                   all        100        271      0.515      0.275      0.224      0.106\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       5/30       1.5G      2.505      1.658      1.056         60        256: 100% ━━━━━━━━━━━━ 8/8 11.8it/s 0.7s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 11.8it/s 0.2s\n",
-      "                   all        100        271      0.445      0.207      0.196     0.0749\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       6/30       1.5G      2.526      1.676      1.044         64        256: 100% ━━━━━━━━━━━━ 8/8 10.2it/s 0.8s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 11.9it/s 0.2s\n",
-      "                   all        100        271      0.822      0.146      0.128     0.0355\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       7/30      1.54G        2.4      1.562      0.999         97        256: 100% ━━━━━━━━━━━━ 8/8 11.9it/s 0.7s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 13.9it/s 0.1s\n",
-      "                   all        100        271      0.513      0.286      0.231     0.0919\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       8/30      1.54G       2.42      1.508      1.004         55        256: 100% ━━━━━━━━━━━━ 8/8 15.1it/s 0.5s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 13.2it/s 0.2s\n",
-      "                   all        100        271       0.53      0.303      0.206     0.0736\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "       9/30      1.54G      2.415      1.494      1.013         82        256: 100% ━━━━━━━━━━━━ 8/8 10.8it/s 0.7s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 12.7it/s 0.2s\n",
-      "                   all        100        271       0.84      0.183      0.153     0.0401\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      10/30      1.54G      2.395      1.488     0.9924         73        256: 100% ━━━━━━━━━━━━ 8/8 12.5it/s 0.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 12.8it/s 0.2s\n",
-      "                   all        100        271      0.856      0.188       0.24     0.0599\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      11/30      1.54G      2.344      1.469      1.002         66        256: 100% ━━━━━━━━━━━━ 8/8 12.8it/s 0.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 12.6it/s 0.2s\n",
-      "                   all        100        271      0.235      0.297      0.234     0.0919\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      12/30      1.54G      2.487      1.468     0.9987         83        256: 100% ━━━━━━━━━━━━ 8/8 12.4it/s 0.6s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 15.0it/s 0.1s\n",
-      "                   all        100        271       0.52      0.176      0.179     0.0476\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      13/30      1.54G      2.391       1.43       1.03        137        256: 100% ━━━━━━━━━━━━ 8/8 6.2it/s 1.3s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 4.9it/s 0.4s\n",
-      "                   all        100        271      0.194      0.374      0.203     0.0596\n",
-      "\n",
-      "      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size\n",
-      "      14/30      1.54G      2.343      1.411      1.012        102        256: 100% ━━━━━━━━━━━━ 8/8 6.1it/s 1.3s\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 10.9it/s 0.2s\n",
-      "                   all        100        271      0.531      0.183      0.195     0.0581\n",
-      "\u001b[34m\u001b[1mEarlyStopping: \u001b[0mTraining stopped early as no improvement observed in last 10 epochs. Best results observed at epoch 4, best model saved as best.pt.\n",
-      "To update EarlyStopping(patience=10) pass a new patience value, i.e. `patience=300` or use `patience=0` to disable EarlyStopping.\n",
-      "\n",
-      "14 epochs completed in 0.008 hours.\n",
-      "Optimizer stripped from /workspace/runs/detect/runs/detect/runs/train/lanov_finetune-2/weights/last.pt, 22.5MB\n",
-      "Optimizer stripped from /workspace/runs/detect/runs/detect/runs/train/lanov_finetune-2/weights/best.pt, 22.5MB\n",
-      "\n",
-      "Validating /workspace/runs/detect/runs/detect/runs/train/lanov_finetune-2/weights/best.pt...\n",
-      "Ultralytics 8.4.40 🚀 Python-3.11.10 torch-2.4.1+cu124 CUDA:0 (NVIDIA GeForce RTX 4090, 24210MiB)\n",
-      "Model summary (fused): 73 layers, 11,127,132 parameters, 0 gradients, 28.4 GFLOPs\n",
-      "                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 2/2 12.2it/s 0.2s\n",
-      "                   all        100        271      0.514      0.275      0.224      0.106\n",
-      "                   car         48        262      0.233      0.576      0.376     0.0984\n",
-      "                   van          5          5          1          0     0.0162    0.00943\n",
-      "                 truck          4          4      0.308       0.25      0.279      0.211\n",
-      "Speed: 0.0ms preprocess, 0.1ms inference, 0.0ms loss, 0.4ms postprocess per image\n",
-      "Results saved to \u001b[1m/workspace/runs/detect/runs/detect/runs/train/lanov_finetune-2\u001b[0m\n",
-      "Lánov fine-tuning hotov. Nejlepší váhy: /workspace/runs/detect/runs/detect/runs/train/lanov_finetune-2/weights/best.pt\n"
-     ]
-    }
-   ],
-   "source": [
-    "import zipfile\n",
-    "from pathlib import Path\n",
-    "\n",
-    "LANOV_ZIP  = Path('dataset_lanov.zip')\n",
-    "LANOV_YAML = Path('lanov/dataset.yaml')\n",
-    "\n",
-    "# Rozbal zip pokud yaml ještě není\n",
-    "if LANOV_ZIP.exists() and not LANOV_YAML.exists():\n",
-    "    print('Rozbaluji dataset_lanov.zip ...')\n",
-    "    with zipfile.ZipFile(LANOV_ZIP) as zf:\n",
-    "        zf.extractall('.')\n",
-    "    print('Hotovo.')\n",
-    "\n",
-    "if not LANOV_YAML.exists():\n",
-    "    print('Lánov dataset nenalezen — fine-tuning přeskočen.')\n",
-    "    print('Spusť prepare_dataset.py nebo nahraj dataset_lanov.zip.')\n",
-    "else:\n",
-    "    print(f'Fine-tuning na: {LANOV_YAML}')\n",
-    "\n",
-    "    lanov_model = YOLO(str(best_weights))  # navazuje na VisDrone trénink\n",
-    "\n",
-    "    lanov_results = lanov_model.train(\n",
-    "        data=str(LANOV_YAML),\n",
-    "        epochs=30,\n",
-    "        imgsz=256,\n",
-    "        batch=32,\n",
-    "        device=DEVICE,\n",
-    "        name='lanov_finetune',\n",
-    "        project='runs/detect/runs/train',\n",
-    "        patience=10,\n",
-    "        save=True,\n",
-    "        val=True,\n",
-    "        plots=True,\n",
-    "        degrees=10.0,\n",
-    "        flipud=0.5,\n",
-    "        fliplr=0.5,\n",
-    "        mosaic=0.5,\n",
-    "        lr0=0.001,   # nižší LR pro fine-tuning\n",
-    "    )\n",
-    "\n",
-    "    best_weights = Path(lanov_results.save_dir) / 'weights' / 'best.pt'\n",
-    "    print(f'Lánov fine-tuning hotov. Nejlepší váhy: {best_weights}')\n"
-   ]
+   "outputs": [],
+   "source": "import zipfile\nimport shutil\nfrom pathlib import Path\n\nLANOV_ZIP  = Path('dataset_lanov.zip')\nLANOV_YAML = Path('lanov/dataset.yaml')\n\n# Rozbal zip pokud yaml ještě není\nif LANOV_ZIP.exists() and not LANOV_YAML.exists():\n    print('Rozbaluji dataset_lanov.zip ...')\n    with zipfile.ZipFile(LANOV_ZIP) as zf:\n        zf.extractall('.')\n    print('Hotovo.')\n\nif not LANOV_YAML.exists():\n    print('Lánov dataset nenalezen — fine-tuning přeskočen.')\n    print('Spusť prepare_dataset.py nebo nahraj dataset_lanov.zip.')\nelse:\n    print(f'Fine-tuning na: {LANOV_YAML}')\n\n    # Záloha modelu před fine-tuningem\n    backup = best_weights.parent / 'best_before_finetune.pt'\n    shutil.copy(best_weights, backup)\n    print(f'Záloha uložena: {backup}')\n\n    lanov_model = YOLO(str(best_weights))  # navazuje na VisDrone trénink\n\n    lanov_results = lanov_model.train(\n        data=str(LANOV_YAML),\n        epochs=30,\n        imgsz=256,\n        batch=32,\n        device=DEVICE,\n        name='lanov_finetune',\n        project=str(Path(\"runs/train\").resolve()),\n        patience=10,\n        save=True,\n        val=True,\n        plots=True,\n        degrees=10.0,\n        flipud=0.5,\n        fliplr=0.5,\n        mosaic=0.5,\n        lr0=0.001,   # nižší LR pro fine-tuning\n    )\n\n    best_weights = Path(lanov_results.save_dir) / 'weights' / 'best.pt'\n    print(f'Lánov fine-tuning hotov. Nejlepší váhy: {best_weights}')"
   },
   {
    "cell_type": "code",
@@ -766,4 +184,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}
\ No newline at end of file
diff --git a/best_before_finetune.pt b/best_before_finetune.pt
new file mode 100644
index 00000000..aafa645d
Binary files /dev/null and b/best_before_finetune.pt differ
diff --git a/best.pt b/best_finetuned.pt
similarity index 99%
rename from best.pt
rename to best_finetuned.pt
index 44423dbf..7e328dcd 100644
Binary files a/best.pt and b/best_finetuned.pt differ
diff --git a/map.html b/map.html
index e449b5fd..5593b592 100644
--- a/map.html
+++ b/map.html
@@ -4,7 +4,8 @@
   <meta charset="UTF-8">
   <meta name="viewport" content="width=device-width, initial-scale=1.0">
   <title>HK Aerial — Detekce vozidel</title>
-  <link rel="stylesheet" href="https://unpkg.com/leaflet@1.9.4/dist/leaflet.css" />
+  <script src="https://unpkg.com/maplibre-gl@4.7.1/dist/maplibre-gl.js"></script>
+  <link href="https://unpkg.com/maplibre-gl@4.7.1/dist/maplibre-gl.css" rel="stylesheet" />
   <style>
     * { margin: 0; padding: 0; box-sizing: border-box; }
     html, body { height: 100%; }
@@ -70,61 +71,93 @@
   </div>
 
   <div class="info-box">
-    Zoom 14–19 &nbsp;|&nbsp; Dlaždice: 256×256 px, z=18
+    WebGL &nbsp;|&nbsp; Zoom 14–19 &nbsp;|&nbsp; Dlaždice: 256×256 px, z=18
   </div>
 
-  <script src="https://unpkg.com/leaflet@1.9.4/dist/leaflet.js"></script>
   <script>
-    const map = L.map('map', {
-      center: [50.208, 15.836],
+    const base = window.location.origin;
+
+    const map = new maplibregl.Map({
+      container: 'map',
+      style: {
+        version: 8,
+        glyphs: 'https://demotiles.maplibre.org/font/{fontstack}/{range}.pbf',
+        sources: {
+          osm: {
+            type: 'raster',
+            tiles: [
+              'https://a.tile.openstreetmap.org/{z}/{x}/{y}.png',
+              'https://b.tile.openstreetmap.org/{z}/{x}/{y}.png',
+              'https://c.tile.openstreetmap.org/{z}/{x}/{y}.png'
+            ],
+            tileSize: 256,
+            attribution: '© OpenStreetMap contributors',
+            maxzoom: 19
+          },
+          tiles: {
+            type: 'raster',
+            tiles: [`${base}/tiles/{z}/{x}/{y}.jpg`],
+            tileSize: 256,
+            minzoom: 13,
+            maxzoom: 18,
+            attribution: 'Letecké snímky HK'
+          },
+          annotated: {
+            type: 'raster',
+            tiles: [`${base}/tiles_annotated/{z}/{x}/{y}.jpg`],
+            tileSize: 256,
+            minzoom: 13,
+            maxzoom: 18,
+            attribution: 'Anotované snímky HK'
+          }
+        },
+        layers: [
+          {
+            id: 'osm-layer',
+            type: 'raster',
+            source: 'osm',
+            layout: { visibility: 'none' }
+          },
+          {
+            id: 'tiles-layer',
+            type: 'raster',
+            source: 'tiles',
+            layout: { visibility: 'none' }
+          },
+          {
+            id: 'annotated-layer',
+            type: 'raster',
+            source: 'annotated',
+            layout: { visibility: 'visible' }
+          }
+        ]
+      },
+      center: [15.836, 50.208],
       zoom: 16,
       minZoom: 13,
       maxZoom: 19
     });
 
-    // Base layers
-    const osm = L.tileLayer('https://{s}.tile.openstreetmap.org/{z}/{x}/{y}.png', {
-      attribution: '© OpenStreetMap contributors',
-      maxZoom: 19
-    });
+    map.addControl(new maplibregl.NavigationControl(), 'bottom-right');
+    map.addControl(new maplibregl.ScaleControl({ unit: 'metric' }), 'bottom-right');
 
-    // Custom tile layers — tiles named as "{z}_{x}_{y}.jpg"
-    const tilesLayer = L.tileLayer('tiles/18_{x}_{y}.jpg', {
-      minNativeZoom: 18,
-      maxNativeZoom: 18,
-      minZoom: 13,
-      maxZoom: 19,
-      tileSize: 256,
-      errorTileUrl: 'data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7',
-      attribution: 'Letecké snímky HK'
-    });
+    const layerIds = {
+      osm: 'osm-layer',
+      tiles: 'tiles-layer',
+      annotated: 'annotated-layer'
+    };
 
-    const annotatedLayer = L.tileLayer('tiles_annotated/18_{x}_{y}.jpg', {
-      minNativeZoom: 18,
-      maxNativeZoom: 18,
-      minZoom: 13,
-      maxZoom: 19,
-      tileSize: 256,
-      errorTileUrl: 'data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7',
-      attribution: 'Anotované snímky HK'
-    });
-
-    // Start with annotated layer
-    annotatedLayer.addTo(map);
-
-    const layers = { osm, tiles: tilesLayer, annotated: annotatedLayer };
     let current = 'annotated';
 
     function setLayer(name) {
       if (name === current) return;
-      map.removeLayer(layers[current]);
-      map.addLayer(layers[name]);
+      map.setLayoutProperty(layerIds[current], 'visibility', 'none');
+      map.setLayoutProperty(layerIds[name], 'visibility', 'visible');
       current = name;
       document.querySelectorAll('.layer-toggle button').forEach(b => b.classList.remove('active'));
       document.getElementById('btn-' + name).classList.add('active');
     }
 
-    // Keyboard shortcut: 1/2/3 to switch layers
     document.addEventListener('keydown', e => {
       if (e.key === '1') setLayer('osm');
       if (e.key === '2') setLayer('tiles');
diff --git a/sample_annotations.png b/sample_annotations.png
deleted file mode 100644
index cbf519ee..00000000
Binary files a/sample_annotations.png and /dev/null differ
diff --git a/sample_detections.png b/sample_detections.png
deleted file mode 100644
index 83f28220..00000000
Binary files a/sample_detections.png and /dev/null differ
diff --git a/serve.py b/serve.py
new file mode 100755
index 00000000..0b66cc0d
--- /dev/null
+++ b/serve.py
@@ -0,0 +1,73 @@
+#!/usr/bin/env python3
+"""Local tile server for HK aerial map. Translates /tiles/{z}/{x}/{y}.jpg -> tiles/18_{x}_{y}.jpg"""
+
+import http.server
+import socketserver
+import os
+import re
+import webbrowser
+import threading
+from pathlib import Path
+
+PORT = 8080
+SCRIPT_DIR = Path(__file__).parent
+TRANSPARENT_GIF = (
+    b'GIF89a\x01\x00\x01\x00\x80\x00\x00\xff\xff\xff\x00\x00\x00'
+    b'!\xf9\x04\x00\x00\x00\x00\x00,\x00\x00\x00\x00\x01\x00\x01'
+    b'\x00\x00\x02\x02D\x01\x00;'
+)
+
+TILE_PATTERN = re.compile(
+    r'^/(tiles(?:_annotated)?)/(\d+)/(\d+)/(\d+)\.(jpg|png|gif)$'
+)
+
+
+class Handler(http.server.SimpleHTTPRequestHandler):
+    def do_GET(self):
+        path = self.path.split('?')[0]
+        m = TILE_PATTERN.match(path)
+        if m:
+            layer, z, x, y, ext = m.groups()
+            # Tiles are stored as 18_x_y.jpg regardless of requested zoom
+            flat = SCRIPT_DIR / layer / f'18_{x}_{y}.jpg'
+            if flat.exists():
+                data = flat.read_bytes()
+                self.send_response(200)
+                self.send_header('Content-Type', 'image/jpeg')
+                self.send_header('Content-Length', len(data))
+                self.send_header('Cache-Control', 'public, max-age=3600')
+                self.end_headers()
+                self.wfile.write(data)
+            else:
+                self.send_response(200)
+                self.send_header('Content-Type', 'image/gif')
+                self.send_header('Content-Length', len(TRANSPARENT_GIF))
+                self.end_headers()
+                self.wfile.write(TRANSPARENT_GIF)
+            return
+
+        super().do_GET()
+
+    def log_message(self, fmt, *args):
+        # Only log non-tile requests so output stays readable
+        if not TILE_PATTERN.match(self.path.split('?')[0]):
+            print(f'  {self.address_string()} {fmt % args}')
+
+
+def main():
+    os.chdir(SCRIPT_DIR)
+
+    with socketserver.TCPServer(('', PORT), Handler) as httpd:
+        httpd.allow_reuse_address = True
+        url = f'http://localhost:{PORT}/map.html'
+        print(f'HK Aerial Map  →  {url}')
+        print('Press Ctrl+C to stop.\n')
+        threading.Timer(0.4, webbrowser.open, args=(url,)).start()
+        try:
+            httpd.serve_forever()
+        except KeyboardInterrupt:
+            print('\nServer stopped.')
+
+
+if __name__ == '__main__':
+    main()
diff --git a/vehicle_counts_chart.png b/vehicle_counts_chart.png
deleted file mode 100644
index a8def748..00000000
Binary files a/vehicle_counts_chart.png and /dev/null differ