keras-hub-nightly 0.22.0.dev202508170419__py3-none-any.whl → 0.24.0.dev202511090424__py3-none-any.whl

keras_hub/src/models/d_fine/d_fine_object_detector_preprocessor.py

@@ -0,0 +1,14 @@
+from keras_hub.src.api_export import keras_hub_export
+from keras_hub.src.models.d_fine.d_fine_backbone import DFineBackbone
+from keras_hub.src.models.d_fine.d_fine_image_converter import (
+    DFineImageConverter,
+)
+from keras_hub.src.models.object_detector_preprocessor import (
+    ObjectDetectorPreprocessor,
+)
+
+
+@keras_hub_export("keras_hub.models.DFineObjectDetectorPreprocessor")
+class DFineObjectDetectorPreprocessor(ObjectDetectorPreprocessor):
+    backbone_cls = DFineBackbone
+    image_converter_cls = DFineImageConverter

keras_hub/src/models/d_fine/d_fine_presets.py

@@ -0,0 +1,155 @@
+# Metadata for loading pretrained model weights.
+backbone_presets = {
+    "dfine_nano_coco": {
+        "metadata": {
+            "description": (
+                "D-FINE Nano model, the smallest variant in the family, "
+                "pretrained on the COCO dataset. Ideal for applications "
+                "where computational resources are limited."
+            ),
+            "params": 3788625,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_nano_coco/1",
+    },
+    "dfine_small_coco": {
+        "metadata": {
+            "description": (
+                "D-FINE Small model pretrained on the COCO dataset. Offers a "
+                "balance between performance and computational efficiency."
+            ),
+            "params": 10329321,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_small_coco/1",
+    },
+    "dfine_medium_coco": {
+        "metadata": {
+            "description": (
+                "D-FINE Medium model pretrained on the COCO dataset. A solid "
+                "baseline with strong performance for general-purpose "
+                "object detection."
+            ),
+            "params": 19621160,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_medium_coco/1",
+    },
+    "dfine_large_coco": {
+        "metadata": {
+            "description": (
+                "D-FINE Large model pretrained on the COCO dataset. Provides "
+                "high accuracy and is suitable for more demanding tasks."
+            ),
+            "params": 31344064,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_large_coco/1",
+    },
+    "dfine_xlarge_coco": {
+        "metadata": {
+            "description": (
+                "D-FINE X-Large model, the largest COCO-pretrained variant, "
+                "designed for state-of-the-art performance where accuracy "
+                "is the top priority."
+            ),
+            "params": 62834048,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_xlarge_coco/1",
+    },
+    "dfine_small_obj365": {
+        "metadata": {
+            "description": (
+                "D-FINE Small model pretrained on the large-scale Objects365 "
+                "dataset, enhancing its ability to recognize a wider "
+                "variety of objects."
+            ),
+            "params": 10623329,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_small_obj365/1",
+    },
+    "dfine_medium_obj365": {
+        "metadata": {
+            "description": (
+                "D-FINE Medium model pretrained on the Objects365 dataset. "
+                "Benefits from a larger and more diverse pretraining corpus."
+            ),
+            "params": 19988670,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_medium_obj365/1",
+    },
+    "dfine_large_obj365": {
+        "metadata": {
+            "description": (
+                "D-FINE Large model pretrained on the Objects365 dataset for "
+                "improved generalization and performance on diverse object "
+                "categories."
+            ),
+            "params": 31858578,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_large_obj365/1",
+    },
+    "dfine_xlarge_obj365": {
+        "metadata": {
+            "description": (
+                "D-FINE X-Large model pretrained on the Objects365 dataset, "
+                "offering maximum performance by leveraging a vast number "
+                "of object categories during pretraining."
+            ),
+            "params": 63348562,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_xlarge_obj365/1",
+    },
+    "dfine_small_obj2coco": {
+        "metadata": {
+            "description": (
+                "D-FINE Small model first pretrained on Objects365 and then "
+                "fine-tuned on COCO, combining broad feature learning with "
+                "benchmark-specific adaptation."
+            ),
+            "params": 10329321,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_small_obj2coco/1",
+    },
+    "dfine_medium_obj2coco": {
+        "metadata": {
+            "description": (
+                "D-FINE Medium model using a two-stage training process: "
+                "pretraining on Objects365 followed by fine-tuning on COCO."
+            ),
+            "params": 19621160,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_medium_obj2coco/1",
+    },
+    "dfine_large_obj2coco_e25": {
+        "metadata": {
+            "description": (
+                "D-FINE Large model pretrained on Objects365 and then "
+                "fine-tuned on COCO for 25 epochs. A high-performance model "
+                "with specialized tuning."
+            ),
+            "params": 31344064,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_large_obj2coco_e25/1",
+    },
+    "dfine_xlarge_obj2coco": {
+        "metadata": {
+            "description": (
+                "D-FINE X-Large model, pretrained on Objects365 and fine-tuned "
+                "on COCO, representing the most powerful model in this "
+                "series for COCO-style tasks."
+            ),
+            "params": 62834048,
+            "path": "d_fine",
+        },
+        "kaggle_handle": "kaggle://keras/d-fine/keras/dfine_xlarge_obj2coco/1",
+    },
+}