Add MobileNetV5 to KerasHub (#2399)

* init: Add initial version

* nit: I like the CI green, ig?

* fix: Bring the numerics significantly closer

* ✅ Yayy, 1e-5 is insanely good!

* chore: Address Gemini reviews

* review #1: Resolve reviews

* feat: Add direct HF / TIMM checkpoint loading

* refactor: Refactor handling of the MobileNetV5 architecture

* nit: Updated examples in backbone and image classifier

* fix: Save backbone and preprocessing, update docstring

Author: harshaljanjani

keras_hub/api/layers/__init__.py

@@ -108,6 +108,9 @@
 from keras_hub.src.models.mobilenet.mobilenet_image_converter import (
     MobileNetImageConverter as MobileNetImageConverter,
 )
+from keras_hub.src.models.mobilenetv5.mobilenetv5_image_converter import (
+    MobileNetV5ImageConverter as MobileNetV5ImageConverter,
+)
 from keras_hub.src.models.moonshine.moonshine_audio_converter import (
     MoonshineAudioConverter as MoonshineAudioConverter,
 )

keras_hub/api/models/__init__.py

@@ -428,6 +428,15 @@
 from keras_hub.src.models.mobilenet.mobilenet_image_classifier_preprocessor import (
     MobileNetImageClassifierPreprocessor as MobileNetImageClassifierPreprocessor,
 )
+from keras_hub.src.models.mobilenetv5.mobilenetv5_backbone import (
+    MobileNetV5Backbone as MobileNetV5Backbone,
+)
+from keras_hub.src.models.mobilenetv5.mobilenetv5_image_classifier import (
+    MobileNetV5ImageClassifier as MobileNetV5ImageClassifier,
+)
+from keras_hub.src.models.mobilenetv5.mobilenetv5_image_classifier_preprocessor import (
+    MobileNetV5ImageClassifierPreprocessor as MobileNetV5ImageClassifierPreprocessor,
+)
 from keras_hub.src.models.moonshine.moonshine_audio_to_text import (
     MoonshineAudioToText as MoonshineAudioToText,
 )