|
import torch |
|
import torch.nn as nn |
|
|
|
from .layers import PatchTransformerEncoder, PixelWiseDotProduct |
|
|
|
|
|
class mViT(nn.Module): |
|
def __init__(self, in_channels, n_query_channels=128, patch_size=16, dim_out=256, |
|
embedding_dim=128, num_heads=4, norm='linear'): |
|
super(mViT, self).__init__() |
|
self.norm = norm |
|
self.n_query_channels = n_query_channels |
|
self.patch_transformer = PatchTransformerEncoder(in_channels, patch_size, embedding_dim, num_heads) |
|
self.dot_product_layer = PixelWiseDotProduct() |
|
|
|
self.conv3x3 = nn.Conv2d(in_channels, embedding_dim, kernel_size=3, stride=1, padding=1) |
|
self.regressor = nn.Sequential(nn.Linear(embedding_dim, 256), |
|
nn.LeakyReLU(), |
|
nn.Linear(256, 256), |
|
nn.LeakyReLU(), |
|
nn.Linear(256, dim_out)) |
|
|
|
def forward(self, x): |
|
|
|
tgt = self.patch_transformer(x.clone()) |
|
|
|
x = self.conv3x3(x) |
|
|
|
regression_head, queries = tgt[0, ...], tgt[1:self.n_query_channels + 1, ...] |
|
|
|
|
|
queries = queries.permute(1, 0, 2) |
|
range_attention_maps = self.dot_product_layer(x, queries) |
|
|
|
y = self.regressor(regression_head) |
|
if self.norm == 'linear': |
|
y = torch.relu(y) |
|
eps = 0.1 |
|
y = y + eps |
|
elif self.norm == 'softmax': |
|
return torch.softmax(y, dim=1), range_attention_maps |
|
else: |
|
y = torch.sigmoid(y) |
|
y = y / y.sum(dim=1, keepdim=True) |
|
return y, range_attention_maps |
|
|