super-gradients/super_gradients/training/models/detection_models/csp_darknet53.py

"""
CSP Darknet

credits: https://github.com/ultralytics
"""
import torch
import torch.nn as nn
from super_gradients.training.utils.utils import get_param, HpmStruct
from super_gradients.training.models.sg_module import SgModule


def autopad(kernel, padding=None):
    # PAD TO 'SAME'
    if padding is None:
        padding = kernel // 2 if isinstance(kernel, int) else [x // 2 for x in kernel]
    return padding


def width_multiplier(original, factor):
    return int(original * factor)


class NumClassesMissingException(Exception):
    pass


class Conv(nn.Module):
    # STANDARD CONVOLUTION
    def __init__(self, input_channels, output_channels, kernel=1, stride=1, padding=None, groups=1,
                 activation_func_type: type = nn.Hardswish):
        super().__init__()

        self.conv = nn.Conv2d(input_channels, output_channels, kernel, stride, autopad(kernel, padding), groups=groups,
                              bias=False)
        self.bn = nn.BatchNorm2d(output_channels)
        self.act = activation_func_type()

    def forward(self, x):
        return self.act(self.bn(self.conv(x)))

    def fuseforward(self, x):
        return self.act(self.conv(x))


class Bottleneck(nn.Module):
    # STANDARD BOTTLENECK
    def __init__(self, input_channels, output_channels, shortcut=True, groups=1,
                 width_mult_factor: float = 1.0,
                 activation_func_type: type = nn.Hardswish):
        super().__init__()

        input_channels = width_multiplier(input_channels, width_mult_factor)
        output_channels = width_multiplier(output_channels, width_mult_factor)
        hidden_channels = output_channels
        self.cv1 = Conv(input_channels, hidden_channels, 1, 1, activation_func_type=activation_func_type)
        self.cv2 = Conv(hidden_channels, output_channels, 3, 1, groups=groups, activation_func_type=activation_func_type)
        self.add = shortcut and input_channels == output_channels

    def forward(self, x):
        return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))


class C3(nn.Module):
    # CSP Bottleneck with 3 convolutions https://github.com/ultralytics/yolov5
    def __init__(self, input_channels, output_channels, bottleneck_blocks_num=1, shortcut=True, groups=1, expansion=0.5,
                 width_mult_factor: float = 1.0, depth_mult_factor: float = 1.0,
                 activation_func_type: type = nn.SiLU):
        super().__init__()

        input_channels = width_multiplier(input_channels, width_mult_factor)
        output_channels = width_multiplier(output_channels, width_mult_factor)
        hidden_channels = int(output_channels * expansion)

        bottleneck_blocks_num = max(round(bottleneck_blocks_num * depth_mult_factor),
                                    1) if bottleneck_blocks_num > 1 else bottleneck_blocks_num

        self.cv1 = Conv(input_channels, hidden_channels, 1, 1, activation_func_type=activation_func_type)
        self.cv2 = Conv(input_channels, hidden_channels, 1, 1, activation_func_type=activation_func_type)
        self.cv3 = Conv(2 * hidden_channels, output_channels, 1, activation_func_type=activation_func_type)
        self.m = nn.Sequential(*[Bottleneck(hidden_channels, hidden_channels, shortcut, groups,
                                            activation_func_type=activation_func_type) for _ in
                                 range(bottleneck_blocks_num)])

    def forward(self, x):
        return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), dim=1))


class BottleneckCSP(nn.Module):
    # CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks
    def __init__(self, input_channels, output_channels, bottleneck_blocks_num=1, shortcut=True, groups=1, expansion=0.5,
                 width_mult_factor: float = 1.0, depth_mult_factor: float = 1.0):
        super().__init__()

        input_channels = width_multiplier(input_channels, width_mult_factor)
        output_channels = width_multiplier(output_channels, width_mult_factor)
        hidden_channels = int(output_channels * expansion)

        bottleneck_blocks_num = max(round(bottleneck_blocks_num * depth_mult_factor),
                                    1) if bottleneck_blocks_num > 1 else bottleneck_blocks_num

        self.cv1 = Conv(input_channels, hidden_channels, 1, 1)
        self.cv2 = nn.Conv2d(input_channels, hidden_channels, 1, 1, bias=False)
        self.cv3 = nn.Conv2d(hidden_channels, hidden_channels, 1, 1, bias=False)
        self.cv4 = Conv(2 * hidden_channels, output_channels, 1, 1)
        self.bn = nn.BatchNorm2d(2 * hidden_channels)  # APPLIED TO CAT(CV2, CV3)
        self.act = nn.LeakyReLU(0.1, inplace=True)
        self.m = nn.Sequential(*[Bottleneck(hidden_channels, hidden_channels, shortcut, groups) for _ in
                                 range(bottleneck_blocks_num)])

    def forward(self, x):
        y1 = self.cv3(self.m(self.cv1(x)))
        y2 = self.cv2(x)
        return self.cv4(self.act(self.bn(torch.cat((y1, y2), dim=1))))


class SPP(nn.Module):
    # SPATIAL PYRAMID POOLING LAYER USED IN YOLOV3-SPP
    def __init__(self, input_channels, output_channels, k=(5, 9, 13), width_mult_factor: float = 1.0):
        super().__init__()

        input_channels = width_multiplier(input_channels, width_mult_factor)
        output_channels = width_multiplier(output_channels, width_mult_factor)
        hidden_channels = input_channels // 2
        self.cv1 = Conv(input_channels, hidden_channels, 1, 1)
        self.cv2 = Conv(hidden_channels * (len(k) + 1), output_channels, 1, 1)
        self.m = nn.ModuleList([nn.MaxPool2d(kernel_size=x, stride=1, padding=x // 2) for x in k])

    def forward(self, x):
        x = self.cv1(x)
        return self.cv2(torch.cat([x] + [m(x) for m in self.m], 1))


class SPPF(nn.Module):
    # Spatial Pyramid Pooling - Fast (SPPF) layer for YOLOv5 by Glenn Jocher https://github.com/ultralytics/yolov5
    # equivalent to SPP(k=(5, 9, 13))
    def __init__(self, input_channels, output_channels, k: int = 5, width_mult_factor: float = 1.0,
                 activation_func_type: type = nn.SiLU):
        super().__init__()

        input_channels = width_multiplier(input_channels, width_mult_factor)
        output_channels = width_multiplier(output_channels, width_mult_factor)
        hidden_channels = input_channels // 2  # hidden channels
        self.cv1 = Conv(input_channels, hidden_channels, 1, 1, activation_func_type=activation_func_type)
        self.cv2 = Conv(hidden_channels * 4, output_channels, 1, 1, activation_func_type=activation_func_type)
        self.maxpool = nn.MaxPool2d(kernel_size=k, stride=1, padding=k // 2)

    def forward(self, x):
        x = self.cv1(x)
        y1 = self.maxpool(x)
        y2 = self.maxpool(y1)
        return self.cv2(torch.cat([x, y1, y2, self.maxpool(y2)], 1))


class Focus(nn.Module):
    # FOCUS WH INFORMATION INTO C-SPACE
    def __init__(self, input_channels, output_channels, kernel=1, stride=1, padding=None, groups=1,
                 width_mult_factor: float = 1.0):
        super().__init__()

        output_channels = width_multiplier(output_channels, width_mult_factor)
        self.conv = Conv(input_channels * 4, output_channels, kernel, stride, padding, groups)

    def forward(self, x):  # x(b,c,w,h) -> y(b,4c,w/2,h/2)
        return self.conv(torch.cat([x[..., ::2, ::2], x[..., 1::2, ::2], x[..., ::2, 1::2], x[..., 1::2, 1::2]], 1))


class ViewModule(nn.Module):
    """
    Returns a reshaped version of the input, to be used in None-Backbone Mode
    """

    def __init__(self, features=1024):
        super(ViewModule, self).__init__()
        self.features = features

    def forward(self, x):
        return x.view(-1, self.features)


class CSPDarknet53(SgModule):
    def __init__(self, arch_params: HpmStruct):
        super().__init__()
        self.num_classes = arch_params.num_classes
        self.backbone_mode = get_param(arch_params, 'backbone_mode', False)
        self.depth_mult_factor = get_param(arch_params, 'depth_mult_factor', 1.)
        self.width_mult_factor = get_param(arch_params, 'width_mult_factor', 1.)
        self.channels_in = get_param(arch_params, 'channels_in', 3)
        self.struct = get_param(arch_params, 'backbone_struct', [3, 9, 9, 3])

        width_mult = lambda channels: width_multiplier(channels, self.width_mult_factor)

        # # THE MODULES LIST IS APPROACHABLE FROM "OUTSIDE THE CLASS - SO WE CAN CHANGE IT'S STRUCTURE"
        self._modules_list = nn.ModuleList()

        # THE MODULES LIST IS APPROACHABLE FROM "OUTSIDE THE CLASS - SO WE CAN CHANGE IT'S STRUCTURE"
        self._modules_list.append(Focus(self.channels_in, 64, 3, width_mult_factor=self.width_mult_factor))  # 0
        self._modules_list.append(Conv(width_mult(64), width_mult(128), 3, 2))  # 1
        self._modules_list.append(
            BottleneckCSP(128, 128, self.struct[0], width_mult_factor=self.width_mult_factor,
                          depth_mult_factor=self.depth_mult_factor))  # 2
        self._modules_list.append(Conv(width_mult(128), width_mult(256), 3, 2))  # 3
        self._modules_list.append(
            BottleneckCSP(256, 256, self.struct[1], width_mult_factor=self.width_mult_factor,
                          depth_mult_factor=self.depth_mult_factor))  # 4
        self._modules_list.append(Conv(width_mult(256), width_mult(512), 3, 2))  # 5
        self._modules_list.append(
            BottleneckCSP(512, 512, self.struct[2], width_mult_factor=self.width_mult_factor,
                          depth_mult_factor=self.depth_mult_factor))  # 6
        self._modules_list.append(Conv(width_mult(512), width_mult(1024), 3, 2))  # 7
        self._modules_list.append(SPP(1024, 1024, k=(5, 9, 13), width_mult_factor=self.width_mult_factor))  # 8
        self._modules_list.append(
            BottleneckCSP(1024, 1024, self.struct[3], False, width_mult_factor=self.width_mult_factor,
                          depth_mult_factor=self.depth_mult_factor))  # 9

        if not self.backbone_mode:
            # IF NOT USED AS A BACKEND BUT AS A CLASSIFIER WE ADD THE CLASSIFICATION LAYERS
            self._modules_list.append(nn.AdaptiveAvgPool2d((1, 1)))
            self._modules_list.append(ViewModule(1024))
            self._modules_list.append(nn.Linear(1024, self.num_classes))

    def forward(self, x):
        return self._modules_list(x)