mblt-vision-python 0.0.0__py3-none-any.whl

mblt_vision/datasets/nyu-depth.yaml

@@ -0,0 +1,7 @@
+name: nyu-depth
+path: ~/.mblt_model_zoo/datasets/nyu-depth
+val: images
+tasks:
+  - depth_estimation
+download:
+  url: https://github.com/ultralytics/assets/releases/download/v0.0.0/nyu-depth.zip

mblt_vision/datasets/registry.py

@@ -0,0 +1,158 @@
+"""Load YAML-backed vision dataset definitions."""
+
+from __future__ import annotations
+
+import copy
+from functools import lru_cache
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+from .._tasks import normalize_vision_task
+
+DATASET_CONFIG_DIR = Path(__file__).parent
+
+
+@lru_cache(maxsize=None)
+def _load_dataset_config(name: str) -> dict[str, Any]:
+    """Load and validate a dataset definition without mutating its cached value."""
+
+    config_path = DATASET_CONFIG_DIR / f"{name.lower()}.yaml"
+    if not config_path.is_file():
+        raise FileNotFoundError(f"Vision dataset definition not found: {config_path}")
+
+    with config_path.open(encoding="utf-8") as config_file:
+        config = yaml.safe_load(config_file)
+    if not isinstance(config, dict):
+        raise ValueError(f"Vision dataset definition must be a mapping: {config_path}")
+    if not isinstance(config.get("path"), str):
+        raise ValueError(
+            f"Vision dataset definition requires a string `path`: {config_path}"
+        )
+    if not isinstance(config.get("name"), str):
+        raise ValueError(
+            f"Vision dataset definition requires a string `name`: {config_path}"
+        )
+    if not isinstance(config.get("tasks"), list) or not all(
+        isinstance(task, str) for task in config["tasks"]
+    ):
+        raise ValueError(
+            f"Vision dataset definition requires a string-list `tasks`: {config_path}"
+        )
+    config["path"] = str(Path(config["path"]).expanduser())
+    return config
+
+
+def get_dataset_config(name: str) -> dict[str, Any]:
+    """Load a named vision dataset definition.
+
+    Args:
+        name: Dataset filename stem, such as ``dotav1``.
+
+    Returns:
+        Parsed dataset configuration with its path expanded.
+
+    Raises:
+        FileNotFoundError: If the dataset definition does not exist.
+        ValueError: If the definition is malformed.
+    """
+
+    return copy.deepcopy(_load_dataset_config(name))
+
+
+def get_dataset_class_names(name: str) -> tuple[str, ...]:
+    """Return ordered class names from a dataset definition.
+
+    Args:
+        name: Dataset filename stem, such as ``coco``.
+
+    Returns:
+        Class names ordered by contiguous zero-based index.
+
+    Raises:
+        ValueError: If the dataset has no valid contiguous ``names`` mapping.
+    """
+
+    names = _load_dataset_config(name).get("names")
+    if not isinstance(names, dict) or not all(
+        isinstance(index, int) and isinstance(label, str)
+        for index, label in names.items()
+    ):
+        raise ValueError(
+            f"Vision dataset definition requires an integer-keyed `names` mapping: {name}"
+        )
+    if set(names) != set(range(len(names))):
+        raise ValueError(
+            f"Vision dataset class IDs must be contiguous and zero-based: {name}"
+        )
+    return tuple(names[index] for index in range(len(names)))
+
+
+@lru_cache(maxsize=None)
+def get_dataset_category_ids(name: str) -> tuple[int, ...]:
+    """Return immutable source category IDs from a dataset definition.
+
+    Args:
+        name: Dataset filename stem, such as ``coco``.
+
+    Returns:
+        Source category IDs ordered by contiguous model-output index.
+
+    Raises:
+        ValueError: If the dataset has no integer ``category_ids`` list.
+    """
+
+    category_ids = _load_dataset_config(name).get("category_ids")
+    if not isinstance(category_ids, list) or not all(
+        isinstance(category_id, int) for category_id in category_ids
+    ):
+        raise ValueError(
+            f"Vision dataset definition requires an integer-list `category_ids`: {name}"
+        )
+    return tuple(category_ids)
+
+
+def get_dataset_config_for_task(
+    task: str, dataset: str | None = None
+) -> dict[str, Any]:
+    """Return the validation dataset definition associated with a vision task.
+
+    Args:
+        task: Vision task name from a model postprocess configuration.
+        dataset: Optional output-taxonomy name. When omitted, the first configured
+            task match is returned for backward compatibility.
+
+    Returns:
+        Matching dataset configuration.
+
+    Raises:
+        ValueError: If no configured dataset supports the task.
+    """
+
+    normalized_task = normalize_vision_task(task)
+    if dataset is not None:
+        try:
+            config = get_dataset_config(dataset)
+        except FileNotFoundError as exc:
+            raise ValueError(
+                f"No vision dataset definition exists for taxonomy `{dataset}`."
+            ) from exc
+        configured_tasks = {
+            normalize_vision_task(configured_task)
+            for configured_task in config["tasks"]
+        }
+        if normalized_task not in configured_tasks:
+            raise ValueError(
+                f"Vision dataset `{dataset}` does not support task `{task}`."
+            )
+        return config
+
+    for config_path in sorted(DATASET_CONFIG_DIR.glob("*.yaml")):
+        config = get_dataset_config(config_path.stem)
+        if normalized_task in {
+            normalize_vision_task(configured_task)
+            for configured_task in config["tasks"]
+        }:
+            return config
+    raise ValueError(f"No vision dataset definition supports task `{task}`.")

mblt_vision/datasets/widerface.yaml

@@ -0,0 +1,9 @@
+# WiderFace validation dataset. Paths are relative to `path` unless absolute.
+name: widerface
+path: ~/.mblt_model_zoo/datasets/widerface
+tasks:
+  - face_detection
+val: images
+download:
+  images: https://huggingface.co/datasets/CUHK-CSE/wider_face/resolve/main/data/WIDER_val.zip
+  annotations: https://huggingface.co/datasets/CUHK-CSE/wider_face/resolve/main/data/wider_face_split.zip

mblt_vision/depth_estimation/__init__.py

@@ -0,0 +1,19 @@
+"""Depth estimation model exports."""
+
+from __future__ import annotations
+
+from .._compat import create_model_class
+
+__all__: list[str] = [
+    "YOLO26lDepth",
+    "YOLO26mDepth",
+    "YOLO26nDepth",
+    "YOLO26sDepth",
+    "YOLO26xDepth",
+]
+
+YOLO26lDepth = create_model_class("YOLO26lDepth", __name__)
+YOLO26mDepth = create_model_class("YOLO26mDepth", __name__)
+YOLO26nDepth = create_model_class("YOLO26nDepth", __name__)
+YOLO26sDepth = create_model_class("YOLO26sDepth", __name__)
+YOLO26xDepth = create_model_class("YOLO26xDepth", __name__)

mblt_vision/face_detection/__init__.py

@@ -0,0 +1,43 @@
+"""Face detection exports for the vision package."""
+
+from __future__ import annotations
+
+from .._compat import create_model_class
+
+__all__: list[str] = [
+    "YOLO11l_face",
+    "YOLO11m_face",
+    "YOLO11n_face",
+    "YOLO11s_face",
+    "YOLO12l_face",
+    "YOLO12m_face",
+    "YOLO12n_face",
+    "YOLO12s_face",
+    "YOLOv10l_face",
+    "YOLOv10m_face",
+    "YOLOv10n_face",
+    "YOLOv10s_face",
+    "YOLOv6m_face",
+    "YOLOv6n_face",
+    "YOLOv8l_face",
+    "YOLOv8m_face",
+    "YOLOv8n_face",
+]
+
+YOLO11l_face = create_model_class("YOLO11l_face", __name__)
+YOLO11m_face = create_model_class("YOLO11m_face", __name__)
+YOLO11n_face = create_model_class("YOLO11n_face", __name__)
+YOLO11s_face = create_model_class("YOLO11s_face", __name__)
+YOLO12l_face = create_model_class("YOLO12l_face", __name__)
+YOLO12m_face = create_model_class("YOLO12m_face", __name__)
+YOLO12n_face = create_model_class("YOLO12n_face", __name__)
+YOLO12s_face = create_model_class("YOLO12s_face", __name__)
+YOLOv10l_face = create_model_class("YOLOv10l_face", __name__)
+YOLOv10m_face = create_model_class("YOLOv10m_face", __name__)
+YOLOv10n_face = create_model_class("YOLOv10n_face", __name__)
+YOLOv10s_face = create_model_class("YOLOv10s_face", __name__)
+YOLOv6m_face = create_model_class("YOLOv6m_face", __name__)
+YOLOv6n_face = create_model_class("YOLOv6n_face", __name__)
+YOLOv8l_face = create_model_class("YOLOv8l_face", __name__)
+YOLOv8m_face = create_model_class("YOLOv8m_face", __name__)
+YOLOv8n_face = create_model_class("YOLOv8n_face", __name__)

mblt_vision/image_classification/__init__.py

@@ -0,0 +1,311 @@
+"""Image classification model exports."""
+
+from __future__ import annotations
+
+from .._compat import create_model_class
+
+__all__: list[str] = [
+    "AlexNet",
+    "CAFormer_B36",
+    "CAFormer_M36",
+    "CAFormer_S18",
+    "CAFormer_S36",
+    "CoAtNet_0_RW_224",
+    "CoAtNet_1_RW_224",
+    "CoAtNet_2_RW_224",
+    "ConvFormer_B36",
+    "ConvFormer_M36",
+    "ConvFormer_S18",
+    "ConvFormer_S36",
+    "ConvNeXt_Base",
+    "ConvNeXt_Large",
+    "ConvNeXt_Small",
+    "ConvNeXt_Tiny",
+    "DeiT_Base_Patch16_224",
+    "DeiT_Base_Patch16_384",
+    "DeiT_Small_Patch16_224",
+    "DeiT_Tiny_Patch16_224",
+    "DeiT3_Base_Patch16_224",
+    "DeiT3_Base_Patch16_384",
+    "DeiT3_Large_Patch16_224",
+    "DeiT3_Large_Patch16_384",
+    "DeiT3_Medium_Patch16_224",
+    "DeiT3_Small_Patch16_224",
+    "DeiT3_Small_Patch16_384",
+    "DenseNet121",
+    "DenseNet161",
+    "DenseNet169",
+    "DenseNet201",
+    "EfficientFormer_L1",
+    "EfficientFormer_L3",
+    "EfficientFormer_L7",
+    "EfficientNet_B0",
+    "EfficientNet_B1",
+    "EfficientNet_B2",
+    "EfficientNet_B3",
+    "EfficientNet_B4",
+    "EfficientNet_B5",
+    "EfficientNet_B6",
+    "EfficientNet_B7",
+    "EfficientNet_V2_L",
+    "EfficientNet_V2_M",
+    "EfficientNet_V2_S",
+    "FlexiViT_Base",
+    "FlexiViT_Large",
+    "FlexiViT_Small",
+    "GoogLeNet",
+    "Inception_V3",
+    "LeViT_Conv_128",
+    "LeViT_Conv_128S",
+    "LeViT_Conv_192",
+    "LeViT_Conv_256",
+    "LeViT_Conv_384",
+    "MNASNet0_5",
+    "MNASNet0_75",
+    "MNASNet1_0",
+    "MNASNet1_3",
+    "MobileNet_V2",
+    "MobileNet_V3_Large",
+    "MobileNet_V3_Small",
+    "RegNet_X_1_6GF",
+    "RegNet_X_3_2GF",
+    "RegNet_X_8GF",
+    "RegNet_X_16GF",
+    "RegNet_X_32GF",
+    "RegNet_X_400MF",
+    "RegNet_X_800MF",
+    "RegNet_Y_1_6GF",
+    "RegNet_Y_3_2GF",
+    "RegNet_Y_8GF",
+    "RegNet_Y_16GF",
+    "RegNet_Y_32GF",
+    "RegNet_Y_400MF",
+    "RegNet_Y_800MF",
+    "RepViT_M0_9",
+    "RepViT_M1",
+    "RepViT_M1_0",
+    "RepViT_M1_1",
+    "RepViT_M1_5",
+    "RepViT_M2",
+    "RepViT_M2_3",
+    "RepViT_M3",
+    "ResNet18",
+    "ResNet34",
+    "ResNet50",
+    "ResNet101",
+    "ResNet152",
+    "ResNeXt50_32x4d",
+    "ResNeXt101_32x8d",
+    "ResNeXt101_64x4d",
+    "ShuffleNet_V2_X0_5",
+    "ShuffleNet_V2_X1_0",
+    "ShuffleNet_V2_X1_5",
+    "ShuffleNet_V2_X2_0",
+    "SqueezeNet1_0",
+    "SqueezeNet1_1",
+    "Swin_B",
+    "Swin_S",
+    "Swin_T",
+    "VGG11",
+    "VGG11_BN",
+    "VGG13",
+    "VGG13_BN",
+    "VGG16",
+    "VGG16_BN",
+    "VGG19",
+    "VGG19_BN",
+    "VisFormer_Small",
+    "VisFormer_Tiny",
+    "ViT_B_16",
+    "ViT_B_32",
+    "ViT_Base_Patch8_224",
+    "ViT_Base_Patch16_224",
+    "ViT_Base_Patch16_384",
+    "ViT_Base_Patch32_224",
+    "ViT_Base_Patch32_384",
+    "ViT_L_16",
+    "ViT_L_32",
+    "ViT_Large_Patch16_224",
+    "ViT_Large_Patch16_384",
+    "ViT_Large_Patch32_384",
+    "ViT_Small_Patch16_224",
+    "ViT_Small_Patch16_384",
+    "ViT_Small_Patch32_224",
+    "ViT_Small_Patch32_384",
+    "ViT_Tiny_Patch16_224",
+    "ViT_Tiny_Patch16_384",
+    "Wide_ResNet50_2",
+    "Wide_ResNet101_2",
+    "YOLO11lCls",
+    "YOLO11mCls",
+    "YOLO11nCls",
+    "YOLO11sCls",
+    "YOLO11xCls",
+    "YOLO26lCls",
+    "YOLO26mCls",
+    "YOLO26nCls",
+    "YOLO26sCls",
+    "YOLO26xCls",
+    "YOLOv5lCls",
+    "YOLOv5mCls",
+    "YOLOv5nCls",
+    "YOLOv5sCls",
+    "YOLOv5xCls",
+    "YOLOv8lCls",
+    "YOLOv8mCls",
+    "YOLOv8nCls",
+    "YOLOv8sCls",
+    "YOLOv8xCls",
+]
+
+AlexNet = create_model_class("AlexNet", __name__)
+CAFormer_B36 = create_model_class("CAFormer_B36", __name__)
+CAFormer_M36 = create_model_class("CAFormer_M36", __name__)
+CAFormer_S18 = create_model_class("CAFormer_S18", __name__)
+CAFormer_S36 = create_model_class("CAFormer_S36", __name__)
+CoAtNet_0_RW_224 = create_model_class("CoAtNet_0_RW_224", __name__)
+CoAtNet_1_RW_224 = create_model_class("CoAtNet_1_RW_224", __name__)
+CoAtNet_2_RW_224 = create_model_class("CoAtNet_2_RW_224", __name__)
+ConvFormer_B36 = create_model_class("ConvFormer_B36", __name__)
+ConvFormer_M36 = create_model_class("ConvFormer_M36", __name__)
+ConvFormer_S18 = create_model_class("ConvFormer_S18", __name__)
+ConvFormer_S36 = create_model_class("ConvFormer_S36", __name__)
+ConvNeXt_Base = create_model_class("ConvNeXt_Base", __name__)
+ConvNeXt_Large = create_model_class("ConvNeXt_Large", __name__)
+ConvNeXt_Small = create_model_class("ConvNeXt_Small", __name__)
+ConvNeXt_Tiny = create_model_class("ConvNeXt_Tiny", __name__)
+DeiT_Base_Patch16_224 = create_model_class("DeiT_Base_Patch16_224", __name__)
+DeiT_Base_Patch16_384 = create_model_class("DeiT_Base_Patch16_384", __name__)
+DeiT_Small_Patch16_224 = create_model_class("DeiT_Small_Patch16_224", __name__)
+DeiT_Tiny_Patch16_224 = create_model_class("DeiT_Tiny_Patch16_224", __name__)
+DeiT3_Base_Patch16_224 = create_model_class("DeiT3_Base_Patch16_224", __name__)
+DeiT3_Base_Patch16_384 = create_model_class("DeiT3_Base_Patch16_384", __name__)
+DeiT3_Large_Patch16_224 = create_model_class("DeiT3_Large_Patch16_224", __name__)
+DeiT3_Large_Patch16_384 = create_model_class("DeiT3_Large_Patch16_384", __name__)
+DeiT3_Medium_Patch16_224 = create_model_class("DeiT3_Medium_Patch16_224", __name__)
+DeiT3_Small_Patch16_224 = create_model_class("DeiT3_Small_Patch16_224", __name__)
+DeiT3_Small_Patch16_384 = create_model_class("DeiT3_Small_Patch16_384", __name__)
+DenseNet121 = create_model_class("DenseNet121", __name__)
+DenseNet161 = create_model_class("DenseNet161", __name__)
+DenseNet169 = create_model_class("DenseNet169", __name__)
+DenseNet201 = create_model_class("DenseNet201", __name__)
+EfficientFormer_L1 = create_model_class("EfficientFormer_L1", __name__)
+EfficientFormer_L3 = create_model_class("EfficientFormer_L3", __name__)
+EfficientFormer_L7 = create_model_class("EfficientFormer_L7", __name__)
+EfficientNet_B0 = create_model_class("EfficientNet_B0", __name__)
+EfficientNet_B1 = create_model_class("EfficientNet_B1", __name__)
+EfficientNet_B2 = create_model_class("EfficientNet_B2", __name__)
+EfficientNet_B3 = create_model_class("EfficientNet_B3", __name__)
+EfficientNet_B4 = create_model_class("EfficientNet_B4", __name__)
+EfficientNet_B5 = create_model_class("EfficientNet_B5", __name__)
+EfficientNet_B6 = create_model_class("EfficientNet_B6", __name__)
+EfficientNet_B7 = create_model_class("EfficientNet_B7", __name__)
+EfficientNet_V2_L = create_model_class("EfficientNet_V2_L", __name__)
+EfficientNet_V2_M = create_model_class("EfficientNet_V2_M", __name__)
+EfficientNet_V2_S = create_model_class("EfficientNet_V2_S", __name__)
+FlexiViT_Base = create_model_class("FlexiViT_Base", __name__)
+FlexiViT_Large = create_model_class("FlexiViT_Large", __name__)
+FlexiViT_Small = create_model_class("FlexiViT_Small", __name__)
+GoogLeNet = create_model_class("GoogLeNet", __name__)
+Inception_V3 = create_model_class("Inception_V3", __name__)
+LeViT_Conv_128 = create_model_class("LeViT_Conv_128", __name__)
+LeViT_Conv_128S = create_model_class("LeViT_Conv_128S", __name__)
+LeViT_Conv_192 = create_model_class("LeViT_Conv_192", __name__)
+LeViT_Conv_256 = create_model_class("LeViT_Conv_256", __name__)
+LeViT_Conv_384 = create_model_class("LeViT_Conv_384", __name__)
+MNASNet0_5 = create_model_class("MNASNet0_5", __name__)
+MNASNet0_75 = create_model_class("MNASNet0_75", __name__)
+MNASNet1_0 = create_model_class("MNASNet1_0", __name__)
+MNASNet1_3 = create_model_class("MNASNet1_3", __name__)
+MobileNet_V2 = create_model_class("MobileNet_V2", __name__)
+MobileNet_V3_Large = create_model_class("MobileNet_V3_Large", __name__)
+MobileNet_V3_Small = create_model_class("MobileNet_V3_Small", __name__)
+RegNet_X_1_6GF = create_model_class("RegNet_X_1_6GF", __name__)
+RegNet_X_3_2GF = create_model_class("RegNet_X_3_2GF", __name__)
+RegNet_X_8GF = create_model_class("RegNet_X_8GF", __name__)
+RegNet_X_16GF = create_model_class("RegNet_X_16GF", __name__)
+RegNet_X_32GF = create_model_class("RegNet_X_32GF", __name__)
+RegNet_X_400MF = create_model_class("RegNet_X_400MF", __name__)
+RegNet_X_800MF = create_model_class("RegNet_X_800MF", __name__)
+RegNet_Y_1_6GF = create_model_class("RegNet_Y_1_6GF", __name__)
+RegNet_Y_3_2GF = create_model_class("RegNet_Y_3_2GF", __name__)
+RegNet_Y_8GF = create_model_class("RegNet_Y_8GF", __name__)
+RegNet_Y_16GF = create_model_class("RegNet_Y_16GF", __name__)
+RegNet_Y_32GF = create_model_class("RegNet_Y_32GF", __name__)
+RegNet_Y_400MF = create_model_class("RegNet_Y_400MF", __name__)
+RegNet_Y_800MF = create_model_class("RegNet_Y_800MF", __name__)
+RepViT_M0_9 = create_model_class("RepViT_M0_9", __name__)
+RepViT_M1 = create_model_class("RepViT_M1", __name__)
+RepViT_M1_0 = create_model_class("RepViT_M1_0", __name__)
+RepViT_M1_1 = create_model_class("RepViT_M1_1", __name__)
+RepViT_M1_5 = create_model_class("RepViT_M1_5", __name__)
+RepViT_M2 = create_model_class("RepViT_M2", __name__)
+RepViT_M2_3 = create_model_class("RepViT_M2_3", __name__)
+RepViT_M3 = create_model_class("RepViT_M3", __name__)
+ResNet18 = create_model_class("ResNet18", __name__)
+ResNet34 = create_model_class("ResNet34", __name__)
+ResNet50 = create_model_class("ResNet50", __name__)
+ResNet101 = create_model_class("ResNet101", __name__)
+ResNet152 = create_model_class("ResNet152", __name__)
+ResNeXt50_32x4d = create_model_class("ResNeXt50_32x4d", __name__)
+ResNeXt101_32x8d = create_model_class("ResNeXt101_32x8d", __name__)
+ResNeXt101_64x4d = create_model_class("ResNeXt101_64x4d", __name__)
+ShuffleNet_V2_X0_5 = create_model_class("ShuffleNet_V2_X0_5", __name__)
+ShuffleNet_V2_X1_0 = create_model_class("ShuffleNet_V2_X1_0", __name__)
+ShuffleNet_V2_X1_5 = create_model_class("ShuffleNet_V2_X1_5", __name__)
+ShuffleNet_V2_X2_0 = create_model_class("ShuffleNet_V2_X2_0", __name__)
+SqueezeNet1_0 = create_model_class("SqueezeNet1_0", __name__)
+SqueezeNet1_1 = create_model_class("SqueezeNet1_1", __name__)
+Swin_B = create_model_class("Swin_B", __name__)
+Swin_S = create_model_class("Swin_S", __name__)
+Swin_T = create_model_class("Swin_T", __name__)
+VGG11 = create_model_class("VGG11", __name__)
+VGG11_BN = create_model_class("VGG11_BN", __name__)
+VGG13 = create_model_class("VGG13", __name__)
+VGG13_BN = create_model_class("VGG13_BN", __name__)
+VGG16 = create_model_class("VGG16", __name__)
+VGG16_BN = create_model_class("VGG16_BN", __name__)
+VGG19 = create_model_class("VGG19", __name__)
+VGG19_BN = create_model_class("VGG19_BN", __name__)
+VisFormer_Small = create_model_class("VisFormer_Small", __name__)
+VisFormer_Tiny = create_model_class("VisFormer_Tiny", __name__)
+ViT_B_16 = create_model_class("ViT_B_16", __name__)
+ViT_B_32 = create_model_class("ViT_B_32", __name__)
+ViT_Base_Patch8_224 = create_model_class("ViT_Base_Patch8_224", __name__)
+ViT_Base_Patch16_224 = create_model_class("ViT_Base_Patch16_224", __name__)
+ViT_Base_Patch16_384 = create_model_class("ViT_Base_Patch16_384", __name__)
+ViT_Base_Patch32_224 = create_model_class("ViT_Base_Patch32_224", __name__)
+ViT_Base_Patch32_384 = create_model_class("ViT_Base_Patch32_384", __name__)
+ViT_L_16 = create_model_class("ViT_L_16", __name__)
+ViT_L_32 = create_model_class("ViT_L_32", __name__)
+ViT_Large_Patch16_224 = create_model_class("ViT_Large_Patch16_224", __name__)
+ViT_Large_Patch16_384 = create_model_class("ViT_Large_Patch16_384", __name__)
+ViT_Large_Patch32_384 = create_model_class("ViT_Large_Patch32_384", __name__)
+ViT_Small_Patch16_224 = create_model_class("ViT_Small_Patch16_224", __name__)
+ViT_Small_Patch16_384 = create_model_class("ViT_Small_Patch16_384", __name__)
+ViT_Small_Patch32_224 = create_model_class("ViT_Small_Patch32_224", __name__)
+ViT_Small_Patch32_384 = create_model_class("ViT_Small_Patch32_384", __name__)
+ViT_Tiny_Patch16_224 = create_model_class("ViT_Tiny_Patch16_224", __name__)
+ViT_Tiny_Patch16_384 = create_model_class("ViT_Tiny_Patch16_384", __name__)
+Wide_ResNet50_2 = create_model_class("Wide_ResNet50_2", __name__)
+Wide_ResNet101_2 = create_model_class("Wide_ResNet101_2", __name__)
+YOLO11lCls = create_model_class("YOLO11lCls", __name__)
+YOLO11mCls = create_model_class("YOLO11mCls", __name__)
+YOLO11nCls = create_model_class("YOLO11nCls", __name__)
+YOLO11sCls = create_model_class("YOLO11sCls", __name__)
+YOLO11xCls = create_model_class("YOLO11xCls", __name__)
+YOLO26lCls = create_model_class("YOLO26lCls", __name__)
+YOLO26mCls = create_model_class("YOLO26mCls", __name__)
+YOLO26nCls = create_model_class("YOLO26nCls", __name__)
+YOLO26sCls = create_model_class("YOLO26sCls", __name__)
+YOLO26xCls = create_model_class("YOLO26xCls", __name__)
+YOLOv5lCls = create_model_class("YOLOv5lCls", __name__)
+YOLOv5mCls = create_model_class("YOLOv5mCls", __name__)
+YOLOv5nCls = create_model_class("YOLOv5nCls", __name__)
+YOLOv5sCls = create_model_class("YOLOv5sCls", __name__)
+YOLOv5xCls = create_model_class("YOLOv5xCls", __name__)
+YOLOv8lCls = create_model_class("YOLOv8lCls", __name__)
+YOLOv8mCls = create_model_class("YOLOv8mCls", __name__)
+YOLOv8nCls = create_model_class("YOLOv8nCls", __name__)
+YOLOv8sCls = create_model_class("YOLOv8sCls", __name__)
+YOLOv8xCls = create_model_class("YOLOv8xCls", __name__)

mblt_vision/instance_segmentation/__init__.py

@@ -0,0 +1,65 @@
+"""Instance segmentation model exports."""
+
+from __future__ import annotations
+
+from .._compat import create_model_class
+
+__all__: list[str] = [
+    "YOLO11lSeg",
+    "YOLO11mSeg",
+    "YOLO11nSeg",
+    "YOLO11sSeg",
+    "YOLO11xSeg",
+    "YOLO12lSeg",
+    "YOLO12mSeg",
+    "YOLO12nSeg",
+    "YOLO12sSeg",
+    "YOLO12xSeg",
+    "YOLO26lSeg",
+    "YOLO26mSeg",
+    "YOLO26nSeg",
+    "YOLO26sSeg",
+    "YOLO26xSeg",
+    "YOLOv5lSeg",
+    "YOLOv5mSeg",
+    "YOLOv5nSeg",
+    "YOLOv5sSeg",
+    "YOLOv5xSeg",
+    "YOLOv8lSeg",
+    "YOLOv8mSeg",
+    "YOLOv8nSeg",
+    "YOLOv8sSeg",
+    "YOLOv8xSeg",
+    "GELANcSeg",
+    "YOLOv9cSeg",
+    "YOLOv9eSeg",
+]
+
+YOLO11lSeg = create_model_class("YOLO11lSeg", __name__)
+YOLO11mSeg = create_model_class("YOLO11mSeg", __name__)
+YOLO11nSeg = create_model_class("YOLO11nSeg", __name__)
+YOLO11sSeg = create_model_class("YOLO11sSeg", __name__)
+YOLO11xSeg = create_model_class("YOLO11xSeg", __name__)
+YOLO12lSeg = create_model_class("YOLO12lSeg", __name__)
+YOLO12mSeg = create_model_class("YOLO12mSeg", __name__)
+YOLO12nSeg = create_model_class("YOLO12nSeg", __name__)
+YOLO12sSeg = create_model_class("YOLO12sSeg", __name__)
+YOLO12xSeg = create_model_class("YOLO12xSeg", __name__)
+YOLO26lSeg = create_model_class("YOLO26lSeg", __name__)
+YOLO26mSeg = create_model_class("YOLO26mSeg", __name__)
+YOLO26nSeg = create_model_class("YOLO26nSeg", __name__)
+YOLO26sSeg = create_model_class("YOLO26sSeg", __name__)
+YOLO26xSeg = create_model_class("YOLO26xSeg", __name__)
+YOLOv5lSeg = create_model_class("YOLOv5lSeg", __name__)
+YOLOv5mSeg = create_model_class("YOLOv5mSeg", __name__)
+YOLOv5nSeg = create_model_class("YOLOv5nSeg", __name__)
+YOLOv5sSeg = create_model_class("YOLOv5sSeg", __name__)
+YOLOv5xSeg = create_model_class("YOLOv5xSeg", __name__)
+YOLOv8lSeg = create_model_class("YOLOv8lSeg", __name__)
+YOLOv8mSeg = create_model_class("YOLOv8mSeg", __name__)
+YOLOv8nSeg = create_model_class("YOLOv8nSeg", __name__)
+YOLOv8sSeg = create_model_class("YOLOv8sSeg", __name__)
+YOLOv8xSeg = create_model_class("YOLOv8xSeg", __name__)
+GELANcSeg = create_model_class("GELANcSeg", __name__)
+YOLOv9cSeg = create_model_class("YOLOv9cSeg", __name__)
+YOLOv9eSeg = create_model_class("YOLOv9eSeg", __name__)

mblt_vision/models/AlexNet.yaml

@@ -0,0 +1,23 @@
+DEFAULT: IMAGENET1K_V1
+IMAGENET1K_V1:
+  file_cfg:
+    repo_id: mobilint/AlexNet
+    filename: alexnet_IMAGENET1K_V1.mxq
+    revision: main
+  pre_cfg:
+    Reader:
+      style: pil
+    Resize:
+      size: 256
+      interpolation: bilinear
+    CenterCrop:
+      size:
+      - 224
+      - 224
+    SetOrder:
+      shape: HWC
+    Normalize:
+      style: torch
+  post_cfg:
+    task: image_classification
+    dataset: imagenet

mblt_vision/models/CAFormer_B36.yaml

@@ -0,0 +1,22 @@
+DEFAULT:
+  file_cfg:
+    repo_id: mobilint/CAFormer_B36
+    filename: caformer_b36.mxq
+    revision: main
+  pre_cfg:
+    Reader:
+      style: pil
+    Resize:
+      size: 224
+      interpolation: bicubic
+    CenterCrop:
+      size:
+      - 224
+      - 224
+    SetOrder:
+      shape: HWC
+    Normalize:
+      style: torch
+  post_cfg:
+    task: image_classification
+    dataset: imagenet