import torch
import torch.nn as nn
import torchvision.models.detection.backbone_utils as backbone_utils
import torchvision.models._utils as _utils
import torch.nn.functional as F
from collections import OrderedDict
from models.net import MobileNetV1 as MobileNetV1
from models.net import FPN as FPN
from models.net import SSH as SSH
class ClassHead(nn.Module):
def __init__(self, inchannels=512, num_anchors=3):
super(ClassHead, self).__init__()
self.num_anchors = num_anchors
self.conv1x1 = nn.Conv2d(inchannels, self.num_anchors*2, kernel_size=(1, 1), stride=1, padding=0)
def forward(self, x):
out = self.conv1x1(x)
out = out.permute(0, 2, 3, 1).contiguous()
return out.view(out.shape[0], -1, 2)
class BboxHead(nn.Module):
def __init__(self, inchannels=512, num_anchors=3):
super(BboxHead, self).__init__()
self.conv1x1 = nn.Conv2d(inchannels, num_anchors*4, kernel_size=(1, 1), stride=1, padding=0)
def forward(self, x):
out = self.conv1x1(x)
out = out.permute(0, 2, 3, 1).contiguous()
return out.view(out.shape[0], -1, 4)
class LandmarkHead(nn.Module):
def __init__(self, inchannels=512, num_anchors=3):
super(LandmarkHead, self).__init__()
self.conv1x1 = nn.Conv2d(inchannels, num_anchors*10, kernel_size=(1, 1), stride=1, padding=0)
def forward(self, x):
out = self.conv1x1(x)
out = out.permute(0, 2, 3, 1).contiguous()
return out.view(out.shape[0], -1, 10)
class RetinaFace(nn.Module):
def __init__(self, cfg=None, phase='train'):
"""
:param cfg: Network related settings.
:param phase: train or test.
"""
super(RetinaFace, self).__init__()
self.phase = phase
backbone = None
if cfg['name'] == 'mobilenet0.25':
backbone = MobileNetV1()
if cfg['pretrain']:
checkpoint = torch.load("./weights/mobilenetV1X0.25_pretrain.tar", map_location=torch.device('cpu'))
from collections import OrderedDict
new_state_dict = OrderedDict()
for k, v in checkpoint['state_dict'].items():
name = k[7:]
new_state_dict[name] = v
backbone.load_state_dict(new_state_dict)
elif cfg['name'] == 'Resnet50':
import torchvision.models as models
backbone = models.resnet50(pretrained=cfg['pretrain'])
self.body = _utils.IntermediateLayerGetter(backbone, cfg['return_layers'])
in_channels_stage2 = cfg['in_channel']
in_channels_list = [
in_channels_stage2 * 2,
in_channels_stage2 * 4,
in_channels_stage2 * 8,
]
out_channels = cfg['out_channel']
self.fpn = FPN(in_channels_list, out_channels)
self.ssh1 = SSH(out_channels, out_channels)
self.ssh2 = SSH(out_channels, out_channels)
self.ssh3 = SSH(out_channels, out_channels)
self.classHead = nn.ModuleList([
ClassHead(cfg['out_channel'], 2),
ClassHead(cfg['out_channel'], 2),
ClassHead(cfg['out_channel'], 2)
])
self.bboxHead = nn.ModuleList([
BboxHead(cfg['out_channel'], 2),
BboxHead(cfg['out_channel'], 2),
BboxHead(cfg['out_channel'], 2)
])
self.landmarkHead = nn.ModuleList([
LandmarkHead(cfg['out_channel'], 2),
LandmarkHead(cfg['out_channel'], 2),
LandmarkHead(cfg['out_channel'], 2)
])
def forward(self, inputs):
out = self.body(inputs)
fpn = self.fpn(out)
feature1 = self.ssh1(fpn[0])
feature2 = self.ssh2(fpn[1])
feature3 = self.ssh3(fpn[2])
features = [feature1, feature2, feature3]
bbox_regressions = torch.cat([self.bboxHead[i](feature) for i, feature in enumerate(features)], dim=1)
classifications = torch.cat([self.classHead[i](feature) for i, feature in enumerate(features)], dim=1)
ldm_regressions = torch.cat([self.landmarkHead[i](feature) for i, feature in enumerate(features)], dim=1)
if self.phase == 'train':
output = (bbox_regressions, classifications, ldm_regressions)
else:
output = (bbox_regressions, F.softmax(classifications, dim=-1), ldm_regressions)
return output