Support tiling

application/backend/app/execution/common/geti_config_converter.py

@@ -156,8 +156,14 @@ class TransformsUpdater:
         "max_shear_degree": "shear",
     }
 
+    # Augmentations that cannot be combined with the tiling pipeline.  Mosaic and
+    # MixUp stitch/blend multiple full images together, which is fundamentally
+    # incompatible with splitting an image into tiles, so they must never be added
+    # to a tiling recipe (see the tile recipe NOTE about the AugmentationScheduler).
+    TILING_INCOMPATIBLE_AUGMENTATIONS: ClassVar[set[str]] = {"mosaic", "mixup"}
+
     @classmethod
-    def update(cls, augmentation_params: dict, config: dict) -> None:  # noqa: C901
+    def update(cls, augmentation_params: dict, config: dict) -> None:  # noqa: C901, PLR0912
         """Update augmentations in the config based on Geti model template.
 
         For each augmentation in augmentation_params:
@@ -186,6 +192,9 @@ def update(cls, augmentation_params: dict, config: dict) -> None:  # noqa: C901
         is_ultralytics = config.get("backend") == "ultralytics"
 
         for aug_name, aug_value in augmentation_params.items():
+            if tiling and aug_name in cls.TILING_INCOMPATIBLE_AUGMENTATIONS:
+                logger.info("Augmentation '%s' is incompatible with the Tiling pipeline and will be skipped", aug_name)
+                continue
             if aug_name not in cls.AUGMENTATION_REGISTRY:
                 if tiling:
                     logger.info("Augmentation '%s' is not applicable in Tiling pipeline", aug_name)
@@ -1047,16 +1056,22 @@ def _update_params(config: dict, param_dict: dict) -> None:
         deim_framework = augmentation_params.pop("deim_framework", None)
         training_parameters = param_dict.get("training", {})
 
-        # Update tiling (always applied regardless of DEIM state)
+        # Update tiling first so downstream augmentation handling can react to it.
         TransformsUpdater.update_tiling(tiling, config)
-
-        # When DEIM is enabled, the AugmentationSchedulerCallback owns the pipeline;
-        # user augmentation overrides must be ignored.
-        deim_enabled = deim_framework is True
+        tile_enabled = bool(config["data"].get("tile_config", {}).get("enable_tiler", False))
+
+        # The DEIM adaptive augmentation scheduling framework (mosaic/mixup based)
+        # is incompatible with tiling. The tile recipe intentionally omits the
+        # AugmentationSchedulerCallback and ships a static tiling-friendly pipeline,
+        # so DEIM must be treated as disabled whenever tiling is enabled, regardless
+        # of the requested value. Otherwise the converter would silently ignore the
+        # user's augmentations (deim=True) or add incompatible ones for a tiling run.
+        deim_enabled = deim_framework is True and not tile_enabled
         if not deim_enabled:
             TransformsUpdater.update(augmentation_params, config)
-            if deim_framework is False:
-                # User explicitly disabled DEIM -> remove the scheduler callback
+            if deim_framework is False or tile_enabled:
+                # DEIM disabled by the user, or forced off because tiling is on ->
+                # remove the scheduler callback and fall back to the static pipeline.
                 GetiConfigConverter._disable_deim_framework(config)
 
         # Update training hyperparameters

application/backend/app/execution/training/getitune_trainer.py

@@ -22,7 +22,7 @@
 
 if TYPE_CHECKING:
     from getitune import TaskType
-    from getitune.config.data import SubsetConfig
+    from getitune.config.data import SubsetConfig, TileConfig
     from getitune.data.dataset.base import VisionDataset
     from getitune.engine.engine import Engine
 
@@ -90,6 +90,7 @@ class DatasetInfo:
     getitune_validation_subset_config: SubsetConfig
     getitune_testing_subset_config: SubsetConfig
     revision_id: UUID
+    tile_config: TileConfig
 
 
 @dataclass(frozen=True)
@@ -243,7 +244,7 @@ def prepare_training_configuration(
             return training_config, getitune_training_config
 
     @step("Prepare Training Dataset", 10)
-    def prepare_training_dataset(
+    def prepare_training_dataset(  # noqa: PLR0915
         self,
         project_id: UUID,
         task: Task,
@@ -258,7 +259,9 @@ def prepare_training_dataset(
         Otherwise, it creates a new dataset from the current items in the database with user-verified annotations.
         """
 
-        from getitune.config.data import SamplerConfig, SubsetConfig
+        from getitune import TaskType
+        from getitune.config.data import SamplerConfig, SubsetConfig, TileConfig
+        from getitune.data.dataset.tile import TileDatasetFactory
         from getitune.data.entity.utils import detect_storage_dtype
         from getitune.data.factory import TransformLibFactory
 
@@ -349,6 +352,30 @@ def build_subset_config(subset_name: str) -> SubsetConfig:
                 transforms=test_subset_config.transforms,  # pyrefly: ignore[missing-attribute,bad-argument-type]
             )
 
+            # Build a TileConfig, and wrap each subset dataset with the tiling factory (if tiling is enabled).
+            tile_cfg_data = getitune_training_config["data"].get("tile_config", {})
+            tile_config = TileConfig(**tile_cfg_data)
+            logger.info("Tiling is set to {}", tile_config.enable_tiler)
+            if tile_config.enable_tiler and getitune_task_type in (TaskType.DETECTION, TaskType.INSTANCE_SEGMENTATION):
+                logger.info("TileConfig for training: {}", tile_config)
+                logger.info("Wrapping subset datasets with TileDatasetFactory")
+                getitune_training_dataset = TileDatasetFactory.create(
+                    dataset=getitune_training_dataset, tile_config=tile_config
+                )
+                getitune_validation_dataset = TileDatasetFactory.create(
+                    dataset=getitune_validation_dataset, tile_config=tile_config
+                )
+                getitune_testing_dataset = TileDatasetFactory.create(
+                    dataset=getitune_testing_dataset, tile_config=tile_config
+                )
+
+            logger.info(
+                "Training dataset size: {}, Validation dataset size: {}, Testing dataset size: {}",
+                len(getitune_training_dataset),
+                len(getitune_validation_dataset),
+                len(getitune_testing_dataset),
+            )
+            logger.info("Augmentations applied {}", getitune_training_config["data"])
             return DatasetInfo(
                 getitune_training_dataset=getitune_training_dataset,
                 getitune_validation_dataset=getitune_validation_dataset,
@@ -357,6 +384,7 @@ def build_subset_config(subset_name: str) -> SubsetConfig:
                 getitune_validation_subset_config=val_subset_config,
                 getitune_testing_subset_config=test_subset_config,
                 revision_id=dataset_revision_id,
+                tile_config=tile_config,
             )
 
     @step("Prepare Model")
@@ -416,6 +444,8 @@ def train_model(
             val_subset=dataset_info.getitune_validation_subset_config,
             test_subset=dataset_info.getitune_testing_subset_config,
         )
+        # Ensure the datamodule (and downstream model) uses the resolved tiling configuration.
+        datamodule.tile_config = dataset_info.tile_config
 
         logger.info("Instantiating model for training (model_id={})", model_id)
         model_cfg = training_config["model"]

application/backend/app/models/training_configuration/augmentation.py

@@ -564,7 +564,9 @@ class AugmentationParameters(BaseModel):
             "training time. When disabled, a simpler static augmentation pipeline "
             "is used instead, allowing direct control over individual augmentations. "
             "WARNING: when the DEIM framework is enabled, all other augmentation "
-            "parameters below are ignored and will have no effect on training."
+            "parameters below are ignored and will have no effect on training. "
+            "Note: the DEIM framework is not supported when Tiling algorithm is "
+            "enabled and will be automatically disabled in that case."
         ),
     )
     random_zoom_out: RandomZoomOut | None = Field(
@@ -588,7 +590,10 @@ class AugmentationParameters(BaseModel):
     mosaic: Mosaic | None = Field(
         default=None,
         title="Mosaic",
-        description="Combines 4 images into one mosaic for augmentation. Applied before resize.",
+        description=(
+            "Combines 4 images into one mosaic for augmentation. Applied before resize. "
+            "Note: this augmentation is not supported when Tiling algorithm is enabled."
+        ),
         json_schema_extra={"depends_on": {"deim_framework": [False, None]}},
     )
     random_resize_crop: RandomResizeCrop | None = Field(
@@ -613,7 +618,10 @@ class AugmentationParameters(BaseModel):
     mixup: Mixup | None = Field(
         default=None,
         title="Mixup",
-        description="Blends two images and their labels for augmentation. Applied before resize.",
+        description=(
+            "Blends two images and their labels for augmentation. Applied before resize. "
+            "Note: this augmentation is not supported when Tiling algorithm is enabled."
+        ),
         json_schema_extra={"depends_on": {"deim_framework": [False, None]}},
     )
     random_horizontal_flip: RandomHorizontalFlip | None = Field(

application/backend/app/supported_models/manifests/detection/dfine_l.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 54.7
     coco_map_50: 72.4
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0005

application/backend/app/supported_models/manifests/detection/dfine_m.yaml

@@ -13,9 +13,6 @@ stats:
   benchmark_metrics:
     coco_map_50_95: 52.7
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0004

application/backend/app/supported_models/manifests/detection/dfine_x.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 56.5
     coco_map_50: 74.0
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0005

application/backend/app/supported_models/manifests/detection/dinov3_detr_l.yaml

@@ -13,9 +13,6 @@ stats:
   benchmark_metrics:
     coco_map_50_95: 56.0
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0005

application/backend/app/supported_models/manifests/detection/dinov3_detr_m.yaml

@@ -13,9 +13,6 @@ stats:
   benchmark_metrics:
     coco_map_50_95: 53.0
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0004

application/backend/app/supported_models/manifests/detection/dinov3_detr_s.yaml

@@ -13,9 +13,6 @@ stats:
   benchmark_metrics:
     coco_map_50_95: 50.9
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0004

application/backend/app/supported_models/manifests/detection/rfdetr_l.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 56.5
     coco_map_50: 75.1
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0001

application/backend/app/supported_models/manifests/detection/rfdetr_m.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 54.7
     coco_map_50: 73.6
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.00021

application/backend/app/supported_models/manifests/detection/rfdetr_n.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 48.4
     coco_map_50: 67.6
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.00021

application/backend/app/supported_models/manifests/detection/rfdetr_s.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 53.0
     coco_map_50: 72.1
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.00021

application/backend/app/supported_models/manifests/instance_segmentation/rfdetr_2xl.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 49.9
     coco_map_50: 73.1
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0001

application/backend/app/supported_models/manifests/instance_segmentation/rfdetr_l.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 47.1
     coco_map_50: 70.5
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0001

application/backend/app/supported_models/manifests/instance_segmentation/rfdetr_m.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 45.3
     coco_map_50: 68.4
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0001

application/backend/app/supported_models/manifests/instance_segmentation/rfdetr_n.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 40.3
     coco_map_50: 63.0
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0001

application/backend/app/supported_models/manifests/instance_segmentation/rfdetr_s.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 43.1
     coco_map_50: 66.2
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0001

application/backend/app/supported_models/manifests/instance_segmentation/rfdetr_xl.yaml

@@ -14,9 +14,6 @@ stats:
     coco_map_50_95: 48.8
     coco_map_50: 72.2
 
-capabilities:
-  tiling: false
-
 hyperparameters:
   training:
     learning_rate: 0.0001

application/backend/tests/integration/services/test_training_configuration_service.py

@@ -148,9 +148,9 @@ def test_get_default_by_model_architecture_strips_tiling_when_unsupported(
         """Tiling parameters should be removed from the default config when the architecture doesn't support tiling."""
         TrainingConfigurationService.get_default_by_model_architecture.cache_clear()
 
-        # D-FINE-M does not support tiling (capabilities.tiling: false in its manifest)
+        # YOLO11-N does not support tiling (capabilities.tiling: false in its manifest)
         result = TrainingConfigurationService.get_default_by_model_architecture(
-            model_architecture_id="object-detection-dfine-m",
+            model_architecture_id="object-detection-yolo11-n",
         )
 
         assert result.algo_level_parameters.dataset_preparation.augmentation.tiling is None
@@ -183,15 +183,15 @@ def test_get_by_model_architecture_strips_tiling_when_unsupported(
         algo_level_config = TrainingConfigurationDB(
             id=str(uuid4()),
             project_id=fxt_project.id,
-            model_architecture_id="object-detection-dfine-m",
+            model_architecture_id="object-detection-yolo11-n",
             configuration_data=algo_level_data,
         )
         db_session.add(algo_level_config)
         db_session.flush()
 
         result = fxt_training_configuration_service.get_by_model_architecture(
             project_id=UUID(fxt_project.id),
-            model_architecture_id="object-detection-dfine-m",
+            model_architecture_id="object-detection-yolo11-n",
         )
 
         assert result.algo_level_parameters.dataset_preparation.augmentation.tiling is None

application/backend/tests/unit/api/schemas/test_training_configuration_view.py

@@ -702,7 +702,8 @@ def fxt_training_configuration_view_json() -> dict:
                                 "key": "mosaic",
                                 "name": "Mosaic",
                                 "description": (
-                                    "Combines 4 images into one mosaic for augmentation. Applied before resize."
+                                    "Combines 4 images into one mosaic for augmentation. Applied before resize. "
+                                    "Note: this augmentation is not supported when Tiling algorithm is enabled."
                                 ),
                                 "depends_on": {"deim_framework": [False, None]},
                                 "parameters": [
@@ -882,7 +883,8 @@ def fxt_training_configuration_view_json() -> dict:
                                 "key": "mixup",
                                 "name": "Mixup",
                                 "description": (
-                                    "Blends two images and their labels for augmentation. Applied before resize."
+                                    "Blends two images and their labels for augmentation. Applied before resize. "
+                                    "Note: this augmentation is not supported when Tiling algorithm is enabled."
                                 ),
                                 "depends_on": {"deim_framework": [False, None]},
                                 "parameters": [