modules/head.py

import torch
import torch.nn as nn
from .blocks.complexblock import ComplexConv2d, ComplexBatchNorm


class CVHead(nn.Module):
    def __init__(self, in_channels, out_channels):
        super(CVHead, self).__init__()
        # self.up = nn.ConvTranspose2d(in_chan, in_chan, kernel_size=(2, 2), stride=(2, 2))
        self.conv_last = ComplexConv2d(in_channels, out_channels, kernel_size=(1, 1))
        self.bn = ComplexBatchNorm(out_channels)

    def forward(self, x):
        out = self.conv_last(x)
        return out

class CVVOCOSHead(nn.Module):
    def __init__(self, in_channels, out_channels, complex_axis=1):
        super(CVVOCOSHead, self).__init__()
        self.real_out = nn.Linear(in_channels, out_channels)
        self.imag_out = nn.Linear(in_channels, out_channels)
        self.complex_axis = complex_axis

    def forward(self, x):
        real, imag = torch.chunk(x, 2, dim=self.complex_axis) # Split real and imaginary parts
        real_out = self.real_out(real.transpose(1, 2)).transpose(1, 2) # Apply linear layer to real part
        imag_out = self.imag_out(imag.transpose(1, 2)).transpose(1, 2) # Apply linear layer to imaginary part
        out = torch.stack([real_out, imag_out], dim=self.complex_axis) # Concatenate real and imaginary parts back together
        return out