def __init__(self, topN=4):
super(attention_net, self).__init__()
self.pretrained_model = resnet.resnet50(pretrained=True)
self.pretrained_model.add_module('raw_fc',
nn.Linear(512 * 4, LABEL_NUM))
self.proposal_net = ProposalNet()
self.topN = topN
self.concat_net = nn.Linear(2048 * (CAT_NUM + 1), LABEL_NUM)
self.partcls_net = nn.Linear(512 * 4, LABEL_NUM)
_, edge_anchors, _ = generate_default_anchor_maps()
self.pad_side = 224
self.edge_anchors = (edge_anchors + 224).astype(np.int)
def __init__(self, topN=4):
super(attention_net, self).__init__()
self.pretrained_model = resnet.resnet50(pretrained=True)
self.pretrained_model.avgpool = nn.AdaptiveAvgPool2d(1)
self.pretrained_model.fc = nn.Linear(512 * 4, 200)
self.proposal_net = ProposalNet()
self.topN = topN
self.concat_net = nn.Linear(2048 * (CAT_NUM + 1), 200)
self.partcls_net = nn.Linear(512 * 4, 200)
_, edge_anchors, _ = generate_default_anchor_maps()
self.pad_side = 224
self.edge_anchors = (edge_anchors + 224).astype(np.int)
def __init__(self, topN=4):
super(attention_net, self).__init__()
self.pretrained_model = resnet.resnet50(pretrained=True)
self.pretrained_model.avgpool = nn.AdaptiveAvgPool2d(1)
self.pretrained_model.fc = nn.Linear(512 * 4, NUM_CLASS)
self.proposal_net = ProposalNet()
self.topN = topN # proposal number
self.concat_net = nn.Linear(2048 * (CAT_NUM + 1), NUM_CLASS)
self.partcls_net = nn.Linear(512 * 4, NUM_CLASS)
#其中每行的4个值[公式]表示矩形左上角和右下角点坐标。9个矩形共有3种形状，3种长宽比
_, edge_anchors, _ = generate_default_anchor_maps()
self.pad_side = 224
self.edge_anchors = (edge_anchors + 224).astype(np.int)
def __init__(self, topN=4):
super(attention_net, self).__init__()
# self.pretrained_model = densenet.densenet201(pretrained=True)
self.pretrained_model = resnet.resnet50(pretrained=True)
self.pretrained_model.avgpool = nn.AdaptiveAvgPool2d(
1) # 1 denotes the output size
num_features = self.pretrained_model.num_features
self.pretrained_model.fc = nn.Linear(
num_features, NUM_CLS) # in_features=512*4, out_features=200
self.proposal_net = ProposalNet(num_features)
self.topN = topN
self.concat_net = nn.Linear(num_features * (CAT_NUM + 1), NUM_CLS)
self.partcls_net = nn.Linear(num_features, NUM_CLS)
_, edge_anchors, _ = generate_default_anchor_maps()
self.pad_side = 224
self.edge_anchors = (edge_anchors + 224).astype(np.int)
def __init__(self, topN=4, n_class=200):
super(attention_net, self).__init__()
self.n_class = n_class
self.pretrained_model = resnet.resnet50(pretrained=True)
self.pretrained_model.avgpool = nn.AdaptiveAvgPool2d(1)
self.pretrained_model.fc = nn.Linear(512 * 4, self.n_class)
self.proposal_net = ProposalNet()
self.topN = topN
self.concat_net = nn.Linear(2048 * (CAT_NUM + 1 + 1), self.n_class)
self.partcls_net = nn.Linear(512 * 4, self.n_class)
self.pad_side = 224
_, edge_anchors_small, _ = generate_default_anchor_maps(
setting='small')
self.edge_anchors_small = (edge_anchors_small + 224).astype(np.int)
_, edge_anchors_large, _ = generate_default_anchor_maps(
setting='large')
self.edge_anchors_large = (edge_anchors_large + 224).astype(np.int)
def __init__(self):
self.pretrained_model = resnet.resnet50(pretrained=True)
self.pretrained_model.avgpool = nn.AdaptiveAvgPool2d(1)
self.pretrained_model.fc = nn.Linear(512 * 4, 200)
trainset = dataset.CUB(root="./CUB_200_2011", is_train=True, data_len=None)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=BATCH_SIZE,
shuffle=True,
num_workers=8,
drop_last=False)
testset = dataset.CUB(root="./CUB_200_2011", is_train=False, data_len=None)
testloader = torch.utils.data.DataLoader(testset,
batch_size=BATCH_SIZE,
shuffle=False,
num_workers=8,
drop_last=False)
# define model
net = resnet.resnet50(pretrained=True)
net.avgpool = torch.nn.AdaptiveAvgPool2d(1)
net.fc = torch.nn.Linear(512 * 4, 200)
if resume:
ckpt = torch.load(resume)
net.load_State_dict(ckpt["net_state_dict"])
start_epoch = ckpt["epoch"] + 1
creterion = torch.nn.CrossEntropyLoss()
# define optimizers
raw_parameters = list(net.parameters())
raw_optimizer = torch.optim.SGD(raw_parameters,
lr=LR,
momentum=0.9,
weight_decay=WD)
