pokouqiege/models/line_detect/line_net.py

import os
from typing import Any, Callable, List, Optional, Tuple, Union
import torch
from torch import nn
from torchvision.ops import MultiScaleRoIAlign

from libs.vision_libs import ops
from libs.vision_libs.models import MobileNet_V3_Large_Weights, mobilenet_v3_large, EfficientNet_V2_S_Weights, \
    efficientnet_v2_s, detection, EfficientNet_V2_L_Weights, efficientnet_v2_l, EfficientNet_V2_M_Weights, \
    efficientnet_v2_m
from libs.vision_libs.models.detection.anchor_utils import AnchorGenerator
from libs.vision_libs.models.detection.rpn import RPNHead, RegionProposalNetwork
from libs.vision_libs.models.detection.ssdlite import _mobilenet_extractor
from libs.vision_libs.models.detection.transform import GeneralizedRCNNTransform
from libs.vision_libs.ops import misc as misc_nn_ops
from libs.vision_libs.transforms._presets import ObjectDetection
from .line_head import LineRCNNHeads
from .line_predictor import LineRCNNPredictor
from libs.vision_libs.models._api import register_model, Weights, WeightsEnum
from libs.vision_libs.models._meta import _COCO_PERSON_CATEGORIES, _COCO_PERSON_KEYPOINT_NAMES, _COCO_CATEGORIES
from libs.vision_libs.models._utils import _ovewrite_value_param, handle_legacy_interface
from libs.vision_libs.models.resnet import resnet50, ResNet50_Weights, ResNet18_Weights, resnet18
from libs.vision_libs.models.detection._utils import overwrite_eps
from libs.vision_libs.models.detection.backbone_utils import _resnet_fpn_extractor, _validate_trainable_layers, \
    BackboneWithFPN
from libs.vision_libs.models.detection.faster_rcnn import FasterRCNN, TwoMLPHead, FastRCNNPredictor

from .roi_heads import RoIHeads
from .trainer import Trainer
from ..base import backbone_factory
from ..base.backbone_factory import get_convnext_fpn, get_anchor_generator
# from ..base.backbone_factory import get_convnext_fpn, get_anchor_generator
from ..base.base_detection_net import BaseDetectionNet
import torch.nn.functional as F

from .predict import Predict1, Predict
from ..base.resnet50fpn import resnet50fpn

from ..config.config_tool import read_yaml

FEATURE_DIM = 8
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

__all__ = [
    "LineNet",
    "LineNet_ResNet50_FPN_Weights",
    "LineNet_ResNet50_FPN_V2_Weights",
    "LineNet_MobileNet_V3_Large_FPN_Weights",
    "LineNet_MobileNet_V3_Large_320_FPN_Weights",
    "linenet_resnet50_fpn",
    "linenet_resnet50_fpn_v2",
    "linenet_mobilenet_v3_large_fpn",
    "linenet_mobilenet_v3_large_320_fpn",
]


def _default_anchorgen():
    anchor_sizes = ((32,), (64,), (128,), (256,), (512,))
    aspect_ratios = ((0.5, 1.0, 2.0),) * len(anchor_sizes)
    return AnchorGenerator(anchor_sizes, aspect_ratios)


class LineNet(BaseDetectionNet):
    # def __init__(self, cfg, **kwargs):
    #     cfg = read_yaml(cfg)
    #     self.cfg=cfg
    #     backbone = cfg['backbone']
    #     print(f'LineNet Backbone:{backbone}')
    #     num_classes = cfg['num_classes']
    #
    #     if backbone == 'resnet50_fpn':
    #         backbone=backbone_factory.get_resnet50_fpn()
    #         print(f'out_chanenels:{backbone.out_channels}')
    #     elif backbone== 'mobilenet_v3_large_fpn':
    #         backbone=backbone_factory.get_mobilenet_v3_large_fpn()
    #     elif backbone=='resnet18_fpn':
    #         backbone=backbone_factory.get_resnet18_fpn()
    #
    #     self.__construct__(backbone=backbone, num_classes=num_classes, **kwargs)

    def __init__(
            self,
            backbone,
            num_classes=None,
            # transform parameters
            min_size=512,
            max_size=1333,
            image_mean=None,
            image_std=None,
            # RPN parameters
            rpn_anchor_generator=None,
            rpn_head=None,
            rpn_pre_nms_top_n_train=2000,
            rpn_pre_nms_top_n_test=1000,
            rpn_post_nms_top_n_train=2000,
            rpn_post_nms_top_n_test=1000,
            rpn_nms_thresh=0.7,
            rpn_fg_iou_thresh=0.7,
            rpn_bg_iou_thresh=0.3,
            rpn_batch_size_per_image=256,
            rpn_positive_fraction=0.5,
            rpn_score_thresh=0.0,
            # Box parameters
            box_roi_pool=None,
            box_head=None,
            box_predictor=None,
            box_score_thresh=0.05,
            box_nms_thresh=0.5,
            box_detections_per_img=100,
            box_fg_iou_thresh=0.5,
            box_bg_iou_thresh=0.5,
            box_batch_size_per_image=512,
            box_positive_fraction=0.25,
            bbox_reg_weights=None,
            # line parameters
            line_head=None,
            line_predictor=None,
            **kwargs,
    ):

        if not hasattr(backbone, "out_channels"):
            raise ValueError(
                "backbone should contain an attribute out_channels "
                "specifying the number of output channels (assumed to be the "
                "same for all the levels)"
            )

        if not isinstance(rpn_anchor_generator, (AnchorGenerator, type(None))):
            raise TypeError(
                f"rpn_anchor_generator should be of type AnchorGenerator or None instead of {type(rpn_anchor_generator)}"
            )
        if not isinstance(box_roi_pool, (MultiScaleRoIAlign, type(None))):
            raise TypeError(
                f"box_roi_pool should be of type MultiScaleRoIAlign or None instead of {type(box_roi_pool)}"
            )

        if num_classes is not None:
            if box_predictor is not None:
                raise ValueError("num_classes should be None when box_predictor is specified")
        else:
            if box_predictor is None:
                raise ValueError("num_classes should not be None when box_predictor is not specified")

        out_channels = backbone.out_channels

        # cfg = read_yaml(cfg)
        # self.cfg=cfg

        if line_head is None:
            num_class = 5
            line_head = LineRCNNHeads(out_channels, num_class)

        if line_predictor is None:
            line_predictor = LineRCNNPredictor()

        if rpn_anchor_generator is None:
            rpn_anchor_generator = _default_anchorgen()
        if rpn_head is None:
            rpn_head = RPNHead(out_channels, rpn_anchor_generator.num_anchors_per_location()[0])

        rpn_pre_nms_top_n = dict(training=rpn_pre_nms_top_n_train, testing=rpn_pre_nms_top_n_test)
        rpn_post_nms_top_n = dict(training=rpn_post_nms_top_n_train, testing=rpn_post_nms_top_n_test)

        rpn = RegionProposalNetwork(
            rpn_anchor_generator,
            rpn_head,
            rpn_fg_iou_thresh,
            rpn_bg_iou_thresh,
            rpn_batch_size_per_image,
            rpn_positive_fraction,
            rpn_pre_nms_top_n,
            rpn_post_nms_top_n,
            rpn_nms_thresh,
            score_thresh=rpn_score_thresh,
        )

        if box_roi_pool is None:
            box_roi_pool = MultiScaleRoIAlign(featmap_names=["0", "1", "2", "3","4"], output_size=7, sampling_ratio=2)

        if box_head is None:
            resolution = box_roi_pool.output_size[0]
            representation_size = 1024
            box_head = TwoMLPHead(out_channels * resolution ** 2, representation_size)

        if box_predictor is None:
            representation_size = 1024
            box_predictor = BoxPredictor(representation_size, num_classes)

        roi_heads = RoIHeads(
            # Box
            box_roi_pool,
            box_head,
            box_predictor,
            line_head,
            line_predictor,
            box_fg_iou_thresh,
            box_bg_iou_thresh,
            box_batch_size_per_image,
            box_positive_fraction,
            bbox_reg_weights,
            box_score_thresh,
            box_nms_thresh,
            box_detections_per_img,
        )

        if image_mean is None:
            image_mean = [0.485, 0.456, 0.406]
        if image_std is None:
            image_std = [0.229, 0.224, 0.225]
        transform = GeneralizedRCNNTransform(min_size, max_size, image_mean, image_std, **kwargs)

        super().__init__(backbone, rpn, roi_heads, transform)

        self.roi_heads = roi_heads

        # self.roi_heads.line_head = line_head
        # self.roi_heads.line_predictor = line_predictor

    def start_train(self, cfg):
        # cfg = read_yaml(cfg)
        self.trainer = Trainer()
        self.trainer.train_from_cfg(model=self, cfg=cfg)

    def load_best_model(self,save_path, device='cuda'):
        if os.path.exists(save_path):
            checkpoint = torch.load(save_path, map_location=device)

            self.load_state_dict(checkpoint['model_state_dict'])
            # if optimizer is not None:
            #     optimizer.load_state_dict(checkpoint['optimizer_state_dict'])
            # epoch = checkpoint['epoch']
            # loss = checkpoint['loss']
            # print(f"Loaded best model from {save_path} at epoch {epoch} with loss {loss:.4f}")
            print(f"Loaded model from {save_path}")
        else:
            print(f"No saved model found at {save_path}")
        return self

    # 加载权重和推理一起
    def predict(self,img_path, type=0, threshold=0.5, save_path=None, show=False):
        # self.predict = Predict(pt_path, model, img_path, type, threshold, save_path, show)
        self.eval()
        self.to(device)
        self.predict = Predict(self, img_path, type, threshold, save_path, show)
        self.predict.run()

    # 不加载权重
    def predict1(self, model, img_path, type=0, threshold=0.5, save_path=None, show=False):
        self.predict = Predict1(model, img_path, type, threshold, save_path, show)
        self.predict.run()


class TwoMLPHead(nn.Module):
    """
    Standard heads for FPN-based models

    Args:
        in_channels (int): number of input channels
        representation_size (int): size of the intermediate representation
    """

    def __init__(self, in_channels, representation_size):
        super().__init__()

        self.fc6 = nn.Linear(in_channels, representation_size)
        self.fc7 = nn.Linear(representation_size, representation_size)

    def forward(self, x):
        x = x.flatten(start_dim=1)

        x = F.relu(self.fc6(x))
        x = F.relu(self.fc7(x))

        return x


class LineNetConvFCHead(nn.Sequential):
    def __init__(
            self,
            input_size: Tuple[int, int, int],
            conv_layers: List[int],
            fc_layers: List[int],
            norm_layer: Optional[Callable[..., nn.Module]] = None,
    ):
        """
        Args:
            input_size (Tuple[int, int, int]): the input size in CHW format.
            conv_layers (list): feature dimensions of each Convolution layer
            fc_layers (list): feature dimensions of each FCN layer
            norm_layer (callable, optional): Module specifying the normalization layer to use. Default: None
        """
        in_channels, in_height, in_width = input_size

        blocks = []
        previous_channels = in_channels
        for current_channels in conv_layers:
            blocks.append(misc_nn_ops.Conv2dNormActivation(previous_channels, current_channels, norm_layer=norm_layer))
            previous_channels = current_channels
        blocks.append(nn.Flatten())
        previous_channels = previous_channels * in_height * in_width
        for current_channels in fc_layers:
            blocks.append(nn.Linear(previous_channels, current_channels))
            blocks.append(nn.ReLU(inplace=True))
            previous_channels = current_channels

        super().__init__(*blocks)
        for layer in self.modules():
            if isinstance(layer, nn.Conv2d):
                nn.init.kaiming_normal_(layer.weight, mode="fan_out", nonlinearity="relu")
                if layer.bias is not None:
                    nn.init.zeros_(layer.bias)


class BoxPredictor(nn.Module):
    """
    Standard classification + bounding box regression layers
    for Fast R-CNN.

    Args:
        in_channels (int): number of input channels
        num_classes (int): number of output classes (including background)
    """

    def __init__(self, in_channels, num_classes):
        super().__init__()
        self.cls_score = nn.Linear(in_channels, num_classes)
        self.bbox_pred = nn.Linear(in_channels, num_classes * 4)

    def forward(self, x):
        if x.dim() == 4:
            torch._assert(
                list(x.shape[2:]) == [1, 1],
                f"x has the wrong shape, expecting the last two dimensions to be [1,1] instead of {list(x.shape[2:])}",
            )
        x = x.flatten(start_dim=1)
        scores = self.cls_score(x)
        bbox_deltas = self.bbox_pred(x)

        return scores, bbox_deltas


_COMMON_META = {
    "categories": _COCO_CATEGORIES,
    "min_size": (1, 1),
}


def create_efficientnetv2_backbone(name='efficientnet_v2_m', pretrained=True):
    # 加载EfficientNetV2模型
    if name == 'efficientnet_v2_s':
        weights = EfficientNet_V2_S_Weights.IMAGENET1K_V1 if pretrained else None
        backbone = efficientnet_v2_s(weights=weights).features
    if name == 'efficientnet_v2_m':
        weights = EfficientNet_V2_M_Weights.IMAGENET1K_V1 if pretrained else None
        backbone = efficientnet_v2_m(weights=weights).features
    if name == 'efficientnet_v2_l':
        weights = EfficientNet_V2_L_Weights.IMAGENET1K_V1 if pretrained else None
        backbone = efficientnet_v2_l(weights=weights).features

    # 定义返回的层索引和名称
    return_layers = {"2": "0", "3": "1", "4": "2", "5": "3"}

    # 获取每个层输出通道数
    in_channels_list = []
    for layer_idx in [2, 3, 4, 5]:
        module = backbone[layer_idx]
        if hasattr(module, 'out_channels'):
            in_channels_list.append(module.out_channels)
        elif hasattr(module[-1], 'out_channels'):
            # 如果module本身没有out_channels，检查最后一个子模块
            in_channels_list.append(module[-1].out_channels)
        else:
            raise ValueError(f"Cannot determine out_channels for layer {layer_idx}")

    # 使用BackboneWithFPN包装backbone
    backbone_with_fpn = BackboneWithFPN(
        backbone=backbone,
        return_layers=return_layers,
        in_channels_list=in_channels_list,
        out_channels=256
    )

    return backbone_with_fpn


def get_line_net_efficientnetv2(num_classes, pretrained_backbone=True):
    # 创建EfficientNetV2 backbone
    backbone = create_efficientnetv2_backbone(pretrained=pretrained_backbone)

    # 确认 backbone 输出特征图数量
    with torch.no_grad():
        images = torch.rand(1,3, 600, 800)
        features = backbone(images)
        featmap_names = list(features.keys())
        print("Feature map names:", featmap_names)  # 例如 ['0', '1', '2', '3']

    # 根据实际特征层数量设置 anchors
    # num_levels = len(featmap_names)
    num_levels=5
    featmap_names= ['0', '1', '2', '3', 'pool']

    anchor_sizes = tuple((int(16 * 2 ** i),) for i in range(num_levels))  # 自动生成不同大小
    print(f'anchor_sizes:{anchor_sizes}')
    aspect_ratios = ((0.5, 1.0, 2.0),) * num_levels  # 所有层共享相同比例
    print(f'aspect_ratios:{aspect_ratios}')

    anchor_generator = AnchorGenerator(
        sizes=anchor_sizes,
        aspect_ratios=aspect_ratios
    )

    # ROI Pooling
    roi_pooler = MultiScaleRoIAlign(
        featmap_names=featmap_names,
        output_size=7,
        sampling_ratio=2
    )

    # 构建模型
    model = LineNet(
        backbone=backbone,
        num_classes=num_classes,
        rpn_anchor_generator=anchor_generator,
        box_roi_pool=roi_pooler
    )

    return model

def get_line_net_convnext_fpn(num_classes=91):
    backbone=get_convnext_fpn()
    featmap_names = ['0', '1', '2', '3', 'pool']
    roi_pooler = MultiScaleRoIAlign(
        featmap_names=featmap_names,
        output_size=7,
        sampling_ratio=2
    )
    test_input = torch.rand(1, 3, 224, 224)
    anchor_generator = get_anchor_generator(backbone, test_input)
    model = LineNet(
        backbone=backbone,
        num_classes=num_classes,  # COCO 数据集有 91 类
        rpn_anchor_generator=anchor_generator,
        box_roi_pool=roi_pooler
    )
    return model

class LineNet_ResNet50_FPN_Weights(WeightsEnum):
    COCO_V1 = Weights(
        url="https://download.pytorch.org/models/fasterrcnn_resnet50_fpn_coco-258fb6c6.pth",
        transforms=ObjectDetection,
        meta={
            **_COMMON_META,
            "num_params": 41755286,
            "recipe": "https://github.com/pytorch/vision/tree/main/references/detection#faster-r-cnn-resnet-50-fpn",
            "_metrics": {
                "COCO-val2017": {
                    "box_map": 37.0,
                }
            },
            "_ops": 134.38,
            "_file_size": 159.743,
            "_docs": """These weights were produced by following a similar training recipe as on the paper.""",
        },
    )
    DEFAULT = COCO_V1


class LineNet_ResNet50_FPN_V2_Weights(WeightsEnum):
    COCO_V1 = Weights(
        url="https://download.pytorch.org/models/fasterrcnn_resnet50_fpn_v2_coco-dd69338a.pth",
        transforms=ObjectDetection,
        meta={
            **_COMMON_META,
            "num_params": 43712278,
            "recipe": "https://github.com/pytorch/vision/pull/5763",
            "_metrics": {
                "COCO-val2017": {
                    "box_map": 46.7,
                }
            },
            "_ops": 280.371,
            "_file_size": 167.104,
            "_docs": """These weights were produced using an enhanced training recipe to boost the model accuracy.""",
        },
    )
    DEFAULT = COCO_V1


class LineNet_MobileNet_V3_Large_FPN_Weights(WeightsEnum):
    COCO_V1 = Weights(
        url="https://download.pytorch.org/models/fasterrcnn_mobilenet_v3_large_fpn-fb6a3cc7.pth",
        transforms=ObjectDetection,
        meta={
            **_COMMON_META,
            "num_params": 19386354,
            "recipe": "https://github.com/pytorch/vision/tree/main/references/detection#faster-r-cnn-mobilenetv3-large-fpn",
            "_metrics": {
                "COCO-val2017": {
                    "box_map": 32.8,
                }
            },
            "_ops": 4.494,
            "_file_size": 74.239,
            "_docs": """These weights were produced by following a similar training recipe as on the paper.""",
        },
    )
    DEFAULT = COCO_V1


class LineNet_MobileNet_V3_Large_320_FPN_Weights(WeightsEnum):
    COCO_V1 = Weights(
        url="https://download.pytorch.org/models/fasterrcnn_mobilenet_v3_large_320_fpn-907ea3f9.pth",
        transforms=ObjectDetection,
        meta={
            **_COMMON_META,
            "num_params": 19386354,
            "recipe": "https://github.com/pytorch/vision/tree/main/references/detection#faster-r-cnn-mobilenetv3-large-320-fpn",
            "_metrics": {
                "COCO-val2017": {
                    "box_map": 22.8,
                }
            },
            "_ops": 0.719,
            "_file_size": 74.239,
            "_docs": """These weights were produced by following a similar training recipe as on the paper.""",
        },
    )
    DEFAULT = COCO_V1


def linenet_newresnet50fpn(
        *,
        weights: Optional[LineNet_ResNet50_FPN_Weights] = None,
        progress: bool = True,
        num_classes: Optional[int] = None,
        weights_backbone: Optional[ResNet18_Weights] = ResNet18_Weights.IMAGENET1K_V1,
        trainable_backbone_layers: Optional[int] = None,
        **kwargs: Any,
) -> LineNet:
    # weights = LineNet_ResNet50_FPN_Weights.verify(weights)
    # weights_backbone = ResNet50_Weights.verify(weights_backbone)

    if weights is not None:
        weights_backbone = None
        num_classes = _ovewrite_value_param("num_classes", num_classes, len(weights.meta["categories"]))
    elif num_classes is None:
        num_classes = 91
    if weights_backbone is not None:
        print(f'resnet50 weights is not None')

    is_trained = weights is not None or weights_backbone is not None
    trainable_backbone_layers = _validate_trainable_layers(is_trained, trainable_backbone_layers, 5, 3)
    norm_layer = misc_nn_ops.FrozenBatchNorm2d if is_trained else nn.BatchNorm2d

    backbone =resnet50fpn()
    featmap_names=['0', '1', '2', '3','pool']
    # print(f'featmap_names:{featmap_names}')
    roi_pooler = MultiScaleRoIAlign(
        featmap_names=featmap_names,
        output_size=7,
        sampling_ratio=2
    )
    num_features=len(featmap_names)
    anchor_sizes = tuple((int(16 * 2 ** i),) for i in range(num_features))  # 自动生成不同大小
    # print(f'anchor_sizes:{anchor_sizes}')
    aspect_ratios = ((0.5, 1.0, 2.0),) * num_features
    # print(f'aspect_ratios:{aspect_ratios}')


    anchor_generator =  AnchorGenerator(sizes=anchor_sizes, aspect_ratios=aspect_ratios)
    # anchors = anchor_generator.generate_anchors()
    # print("Number of anchor sizes:", len(anchor_generator.sizes))  # 应为 5
    model = LineNet(backbone, num_classes=num_classes,anchor_generator=anchor_generator,

                    box_roi_pool=roi_pooler,
                    **kwargs)

    if weights is not None:
        model.load_state_dict(weights.get_state_dict(progress=progress, check_hash=True))
        if weights == LineNet_ResNet50_FPN_Weights.COCO_V1:
            overwrite_eps(model, 0.0)

    return model

# @register_model()
# @handle_legacy_interface(
#     weights=("pretrained", LineNet_ResNet50_FPN_Weights.COCO_V1),
#     weights_backbone=("pretrained_backbone", ResNet50_Weights.IMAGENET1K_V1),
# )
def linenet_resnet18_fpn(
        *,
        weights: Optional[LineNet_ResNet50_FPN_Weights] = None,
        progress: bool = True,
        num_classes: Optional[int] = None,
        weights_backbone: Optional[ResNet18_Weights] = ResNet18_Weights.IMAGENET1K_V1,
        trainable_backbone_layers: Optional[int] = None,
        **kwargs: Any,
) -> LineNet:
    # weights = LineNet_ResNet50_FPN_Weights.verify(weights)
    # weights_backbone = ResNet50_Weights.verify(weights_backbone)

    if weights is not None:
        weights_backbone = None
        num_classes = _ovewrite_value_param("num_classes", num_classes, len(weights.meta["categories"]))
    elif num_classes is None:
        num_classes = 91
    if weights_backbone is not None:
        print(f'resnet50 weights is not None')

    is_trained = weights is not None or weights_backbone is not None
    trainable_backbone_layers = _validate_trainable_layers(is_trained, trainable_backbone_layers, 5, 3)
    norm_layer = misc_nn_ops.FrozenBatchNorm2d if is_trained else nn.BatchNorm2d

    backbone = resnet18(weights=weights_backbone, progress=progress, norm_layer=norm_layer)
    backbone = _resnet_fpn_extractor(backbone, trainable_backbone_layers)
    model = LineNet(backbone, num_classes=num_classes, **kwargs)

    if weights is not None:
        model.load_state_dict(weights.get_state_dict(progress=progress, check_hash=True))
        if weights == LineNet_ResNet50_FPN_Weights.COCO_V1:
            overwrite_eps(model, 0.0)

    return model


def linenet_resnet50_fpn(
        *,
        weights: Optional[LineNet_ResNet50_FPN_Weights] = None,
        progress: bool = True,
        num_classes: Optional[int] = None,
        weights_backbone: Optional[ResNet50_Weights] = ResNet50_Weights.IMAGENET1K_V1,
        trainable_backbone_layers: Optional[int] = None,
        **kwargs: Any,
) -> LineNet:
    """
    Faster R-CNN model with a ResNet-50-FPN backbone from the `Faster R-CNN: Towards Real-Time Object
    Detection with Region Proposal Networks <https://arxiv.org/abs/1506.01497>`__
    paper.

    .. betastatus:: detection module

    The input to the model is expected to be a list of tensors, each of shape ``[C, H, W]``, one for each
    image, and should be in ``0-1`` range. Different images can have different sizes.

    The behavior of the model changes depending on if it is in training or evaluation mode.

    During training, the model expects both the input tensors and a targets (list of dictionary),
    containing:

        - boxes (``FloatTensor[N, 4]``): the ground-truth boxes in ``[x1, y1, x2, y2]`` format, with
          ``0 <= x1 < x2 <= W`` and ``0 <= y1 < y2 <= H``.
        - labels (``Int64Tensor[N]``): the class label for each ground-truth box

    The model returns a ``Dict[Tensor]`` during training, containing the classification and regression
    losses for both the RPN and the R-CNN.

    During inference, the model requires only the input tensors, and returns the post-processed
    predictions as a ``List[Dict[Tensor]]``, one for each input image. The fields of the ``Dict`` are as
    follows, where ``N`` is the number of detections:

        - boxes (``FloatTensor[N, 4]``): the predicted boxes in ``[x1, y1, x2, y2]`` format, with
          ``0 <= x1 < x2 <= W`` and ``0 <= y1 < y2 <= H``.
        - labels (``Int64Tensor[N]``): the predicted labels for each detection
        - scores (``Tensor[N]``): the scores of each detection

    For more details on the output, you may refer to :ref:`instance_seg_output`.

    Faster R-CNN is exportable to ONNX for a fixed batch size with inputs images of fixed size.

    Example::

        >>> model = torchvision.models.detection.fasterrcnn_resnet50_fpn(weights=FasterRCNN_ResNet50_FPN_Weights.DEFAULT)
        >>> # For training
        >>> images, boxes = torch.rand(4, 3, 600, 1200), torch.rand(4, 11, 4)
        >>> boxes[:, :, 2:4] = boxes[:, :, 0:2] + boxes[:, :, 2:4]
        >>> labels = torch.randint(1, 91, (4, 11))
        >>> images = list(image for image in images)
        >>> targets = []
        >>> for i in range(len(images)):
        >>>     d = {}
        >>>     d['boxes'] = boxes[i]
        >>>     d['labels'] = labels[i]
        >>>     targets.append(d)
        >>> output = model(images, targets)
        >>> # For inference
        >>> model.eval()
        >>> x = [torch.rand(3, 300, 400), torch.rand(3, 500, 400)]
        >>> predictions = model(x)
        >>>
        >>> # optionally, if you want to export the model to ONNX:
        >>> torch.onnx.export(model, x, "faster_rcnn.onnx", opset_version = 11)

    Args:
        weights (:class:`~torchvision.models.detection.FasterRCNN_ResNet50_FPN_Weights`, optional): The
            pretrained weights to use. See
            :class:`~torchvision.models.detection.FasterRCNN_ResNet50_FPN_Weights` below for
            more details, and possible values. By default, no pre-trained
            weights are used.
        progress (bool, optional): If True, displays a progress bar of the
            download to stderr. Default is True.
        num_classes (int, optional): number of output classes of the model (including the background)
        weights_backbone (:class:`~torchvision.models.ResNet50_Weights`, optional): The
            pretrained weights for the backbone.
        trainable_backbone_layers (int, optional): number of trainable (not frozen) layers starting from
            final block. Valid values are between 0 and 5, with 5 meaning all backbone layers are
            trainable. If ``None`` is passed (the default) this value is set to 3.
        **kwargs: parameters passed to the ``torchvision.models.detection.faster_rcnn.FasterRCNN``
            base class. Please refer to the `source code
            <https://github.com/pytorch/vision/blob/main/torchvision/models/detection/faster_rcnn.py>`_
            for more details about this class.

    .. autoclass:: torchvision.models.detection.FasterRCNN_ResNet50_FPN_Weights
        :members:
    """
    weights = LineNet_ResNet50_FPN_Weights.verify(weights)
    weights_backbone = ResNet50_Weights.verify(weights_backbone)

    if weights is not None:
        weights_backbone = None
        num_classes = _ovewrite_value_param("num_classes", num_classes, len(weights.meta["categories"]))
    elif num_classes is None:
        num_classes = 91
    if weights_backbone is not None:
        print(f'resnet50 weights is not None')

    is_trained = weights is not None or weights_backbone is not None
    trainable_backbone_layers = _validate_trainable_layers(is_trained, trainable_backbone_layers, 5, 3)
    norm_layer = misc_nn_ops.FrozenBatchNorm2d if is_trained else nn.BatchNorm2d

    backbone = resnet50(weights=weights_backbone, progress=progress, norm_layer=norm_layer)
    backbone = _resnet_fpn_extractor(backbone, trainable_backbone_layers)
    model = LineNet(backbone, num_classes=num_classes, **kwargs)

    if weights is not None:
        model.load_state_dict(weights.get_state_dict(progress=progress, check_hash=True))
        if weights == LineNet_ResNet50_FPN_Weights.COCO_V1:
            overwrite_eps(model, 0.0)

    return model




# @register_model()
# @handle_legacy_interface(
#     weights=("pretrained", LineNet_ResNet50_FPN_V2_Weights.COCO_V1),
#     weights_backbone=("pretrained_backbone", ResNet50_Weights.IMAGENET1K_V1),
# )
def linenet_resnet50_fpn_v2(
        *,
        weights: Optional[LineNet_ResNet50_FPN_V2_Weights] = None,
        progress: bool = True,
        num_classes: Optional[int] = None,
        weights_backbone: Optional[ResNet50_Weights] = None,
        trainable_backbone_layers: Optional[int] = None,
        **kwargs: Any,
) -> LineNet:
    """
    Constructs an improved Faster R-CNN model with a ResNet-50-FPN backbone from `Benchmarking Detection
    Transfer Learning with Vision Transformers <https://arxiv.org/abs/2111.11429>`__ paper.

    .. betastatus:: detection module

    It works similarly to Faster R-CNN with ResNet-50 FPN backbone. See
    :func:`~torchvision.models.detection.fasterrcnn_resnet50_fpn` for more
    details.

    Args:
        weights (:class:`~torchvision.models.detection.FasterRCNN_ResNet50_FPN_V2_Weights`, optional): The
            pretrained weights to use. See
            :class:`~torchvision.models.detection.FasterRCNN_ResNet50_FPN_V2_Weights` below for
            more details, and possible values. By default, no pre-trained
            weights are used.
        progress (bool, optional): If True, displays a progress bar of the
            download to stderr. Default is True.
        num_classes (int, optional): number of output classes of the model (including the background)
        weights_backbone (:class:`~torchvision.models.ResNet50_Weights`, optional): The
            pretrained weights for the backbone.
        trainable_backbone_layers (int, optional): number of trainable (not frozen) layers starting from
            final block. Valid values are between 0 and 5, with 5 meaning all backbone layers are
            trainable. If ``None`` is passed (the default) this value is set to 3.
        **kwargs: parameters passed to the ``torchvision.models.detection.faster_rcnn.FasterRCNN``
            base class. Please refer to the `source code
            <https://github.com/pytorch/vision/blob/main/torchvision/models/detection/faster_rcnn.py>`_
            for more details about this class.

    .. autoclass:: torchvision.models.detection.FasterRCNN_ResNet50_FPN_V2_Weights
        :members:
    """
    weights = LineNet_ResNet50_FPN_V2_Weights.verify(weights)
    weights_backbone = ResNet50_Weights.verify(weights_backbone)

    if weights is not None:
        weights_backbone = None
        num_classes = _ovewrite_value_param("num_classes", num_classes, len(weights.meta["categories"]))
    elif num_classes is None:
        num_classes = 91

    is_trained = weights is not None or weights_backbone is not None
    trainable_backbone_layers = _validate_trainable_layers(is_trained, trainable_backbone_layers, 5, 3)

    backbone = resnet50(weights=weights_backbone, progress=progress)
    backbone = _resnet_fpn_extractor(backbone, trainable_backbone_layers, norm_layer=nn.BatchNorm2d)
    rpn_anchor_generator = _default_anchorgen()
    rpn_head = RPNHead(backbone.out_channels, rpn_anchor_generator.num_anchors_per_location()[0], conv_depth=2)
    box_head = LineNetConvFCHead(
        (backbone.out_channels, 7, 7), [256, 256, 256, 256], [1024], norm_layer=nn.BatchNorm2d
    )
    model = LineNet(
        backbone,
        num_classes=num_classes,
        rpn_anchor_generator=rpn_anchor_generator,
        rpn_head=rpn_head,
        box_head=box_head,
        **kwargs,
    )

    if weights is not None:
        model.load_state_dict(weights.get_state_dict(progress=progress, check_hash=True))

    return model


def _linenet_mobilenet_v3_large_fpn(
        *,
        weights: Optional[Union[LineNet_MobileNet_V3_Large_FPN_Weights, LineNet_MobileNet_V3_Large_320_FPN_Weights]],
        progress: bool,
        num_classes: Optional[int],
        weights_backbone: Optional[MobileNet_V3_Large_Weights],
        trainable_backbone_layers: Optional[int],
        **kwargs: Any,
) -> LineNet:
    if weights is not None:
        weights_backbone = None
        num_classes = _ovewrite_value_param("num_classes", num_classes, len(weights.meta["categories"]))
    elif num_classes is None:
        num_classes = 91

    is_trained = weights is not None or weights_backbone is not None
    trainable_backbone_layers = _validate_trainable_layers(is_trained, trainable_backbone_layers, 6, 3)
    norm_layer = misc_nn_ops.FrozenBatchNorm2d if is_trained else nn.BatchNorm2d

    backbone = mobilenet_v3_large(weights=weights_backbone, progress=progress, norm_layer=norm_layer)
    backbone = _mobilenet_extractor(backbone, True, trainable_backbone_layers)
    anchor_sizes = (
                       (
                           32,
                           64,
                           128,
                           256,
                           512,
                       ),
                   ) * 3
    aspect_ratios = ((0.5, 1.0, 2.0),) * len(anchor_sizes)
    model = LineNet(
        backbone, num_classes, rpn_anchor_generator=AnchorGenerator(anchor_sizes, aspect_ratios), **kwargs
    )

    if weights is not None:
        model.load_state_dict(weights.get_state_dict(progress=progress, check_hash=True))

    return model


# @register_model()
# @handle_legacy_interface(
#     weights=("pretrained", LineNet_MobileNet_V3_Large_320_FPN_Weights.COCO_V1),
#     weights_backbone=("pretrained_backbone", MobileNet_V3_Large_Weights.IMAGENET1K_V1),
# )
def linenet_mobilenet_v3_large_320_fpn(
        *,
        weights: Optional[LineNet_MobileNet_V3_Large_320_FPN_Weights] = None,
        progress: bool = True,
        num_classes: Optional[int] = None,
        weights_backbone: Optional[MobileNet_V3_Large_Weights] = MobileNet_V3_Large_Weights.IMAGENET1K_V1,
        trainable_backbone_layers: Optional[int] = None,
        **kwargs: Any,
) -> LineNet:
    """
    Low resolution Faster R-CNN model with a MobileNetV3-Large backbone tuned for mobile use cases.

    .. betastatus:: detection module

    It works similarly to Faster R-CNN with ResNet-50 FPN backbone. See
    :func:`~torchvision.models.detection.fasterrcnn_resnet50_fpn` for more
    details.

    Example::

        >>> model = torchvision.models.detection.fasterrcnn_mobilenet_v3_large_320_fpn(weights=FasterRCNN_MobileNet_V3_Large_320_FPN_Weights.DEFAULT)
        >>> model.eval()
        >>> x = [torch.rand(3, 300, 400), torch.rand(3, 500, 400)]
        >>> predictions = model(x)

    Args:
        weights (:class:`~torchvision.models.detection.FasterRCNN_MobileNet_V3_Large_320_FPN_Weights`, optional): The
            pretrained weights to use. See
            :class:`~torchvision.models.detection.FasterRCNN_MobileNet_V3_Large_320_FPN_Weights` below for
            more details, and possible values. By default, no pre-trained
            weights are used.
        progress (bool, optional): If True, displays a progress bar of the
            download to stderr. Default is True.
        num_classes (int, optional): number of output classes of the model (including the background)
        weights_backbone (:class:`~torchvision.models.MobileNet_V3_Large_Weights`, optional): The
            pretrained weights for the backbone.
        trainable_backbone_layers (int, optional): number of trainable (not frozen) layers starting from
            final block. Valid values are between 0 and 6, with 6 meaning all backbone layers are
            trainable. If ``None`` is passed (the default) this value is set to 3.
        **kwargs: parameters passed to the ``torchvision.models.detection.faster_rcnn.FasterRCNN``
            base class. Please refer to the `source code
            <https://github.com/pytorch/vision/blob/main/torchvision/models/detection/faster_rcnn.py>`_
            for more details about this class.

    .. autoclass:: torchvision.models.detection.FasterRCNN_MobileNet_V3_Large_320_FPN_Weights
        :members:
    """
    weights = LineNet_MobileNet_V3_Large_320_FPN_Weights.verify(weights)
    weights_backbone = MobileNet_V3_Large_Weights.verify(weights_backbone)

    defaults = {
        "min_size": 320,
        "max_size": 640,
        "rpn_pre_nms_top_n_test": 150,
        "rpn_post_nms_top_n_test": 150,
        "rpn_score_thresh": 0.05,
    }

    kwargs = {**defaults, **kwargs}
    return _linenet_mobilenet_v3_large_fpn(
        weights=weights,
        progress=progress,
        num_classes=num_classes,
        weights_backbone=weights_backbone,
        trainable_backbone_layers=trainable_backbone_layers,
        **kwargs,
    )


# @register_model()
# @handle_legacy_interface(
#     weights=("pretrained", LineNet_MobileNet_V3_Large_FPN_Weights.COCO_V1),
#     weights_backbone=("pretrained_backbone", MobileNet_V3_Large_Weights.IMAGENET1K_V1),
# )
def linenet_mobilenet_v3_large_fpn(
        *,
        weights: Optional[LineNet_MobileNet_V3_Large_FPN_Weights] = None,
        progress: bool = True,
        num_classes: Optional[int] = None,
        weights_backbone: Optional[MobileNet_V3_Large_Weights] = MobileNet_V3_Large_Weights.IMAGENET1K_V1,
        trainable_backbone_layers: Optional[int] = None,
        **kwargs: Any,
) -> LineNet:
    """
    Constructs a high resolution Faster R-CNN model with a MobileNetV3-Large FPN backbone.

    .. betastatus:: detection module

    It works similarly to Faster R-CNN with ResNet-50 FPN backbone. See
    :func:`~torchvision.models.detection.fasterrcnn_resnet50_fpn` for more
    details.

    Example::

        >>> model = torchvision.models.detection.fasterrcnn_mobilenet_v3_large_fpn(weights=FasterRCNN_MobileNet_V3_Large_FPN_Weights.DEFAULT)
        >>> model.eval()
        >>> x = [torch.rand(3, 300, 400), torch.rand(3, 500, 400)]
        >>> predictions = model(x)

    Args:
        weights (:class:`~torchvision.models.detection.FasterRCNN_MobileNet_V3_Large_FPN_Weights`, optional): The
            pretrained weights to use. See
            :class:`~torchvision.models.detection.FasterRCNN_MobileNet_V3_Large_FPN_Weights` below for
            more details, and possible values. By default, no pre-trained
            weights are used.
        progress (bool, optional): If True, displays a progress bar of the
            download to stderr. Default is True.
        num_classes (int, optional): number of output classes of the model (including the background)
        weights_backbone (:class:`~torchvision.models.MobileNet_V3_Large_Weights`, optional): The
            pretrained weights for the backbone.
        trainable_backbone_layers (int, optional): number of trainable (not frozen) layers starting from
            final block. Valid values are between 0 and 6, with 6 meaning all backbone layers are
            trainable. If ``None`` is passed (the default) this value is set to 3.
        **kwargs: parameters passed to the ``torchvision.models.detection.faster_rcnn.FasterRCNN``
            base class. Please refer to the `source code
            <https://github.com/pytorch/vision/blob/main/torchvision/models/detection/faster_rcnn.py>`_
            for more details about this class.

    .. autoclass:: torchvision.models.detection.FasterRCNN_MobileNet_V3_Large_FPN_Weights
        :members:
    """
    weights = LineNet_MobileNet_V3_Large_FPN_Weights.verify(weights)
    weights_backbone = MobileNet_V3_Large_Weights.verify(weights_backbone)

    defaults = {
        "rpn_score_thresh": 0.05,
    }

    kwargs = {**defaults, **kwargs}
    return _linenet_mobilenet_v3_large_fpn(
        weights=weights,
        progress=progress,
        num_classes=num_classes,
        weights_backbone=weights_backbone,
        trainable_backbone_layers=trainable_backbone_layers,
        **kwargs,
    )