train.py

from nets.ssd import get_ssd
from nets.ssd_training import Generator,MultiBoxLoss
from torch.utils.data import DataLoader
from utils.dataloader import ssd_dataset_collate, SSDDataset
from utils.config import Config
from torchsummary import summary
from torch.autograd import Variable
from tqdm import tqdm
import torch.backends.cudnn as cudnn
import time
import torch
import numpy as np
import torch.nn as nn
import torch.optim as optim
import torch.nn.init as init

def get_lr(optimizer):
    for param_group in optimizer.param_groups:
        return param_group['lr']
        
if __name__ == "__main__":
    # ------------------------------------#
    #   先冻结一部分权重训练
    #   后解冻全部权重训练
    #   先大学习率
    #   后小学习率
    # ------------------------------------#
    lr = 5e-4
    freeze_lr = 1e-4
    Cuda = True

    Start_iter = 0
    Freeze_epoch = 25
    Epoch = 50

    Batch_size = 4
    #-------------------------------#
    #   Dataloder的使用
    #-------------------------------#
    Use_Data_Loader = True

    model = get_ssd("train",Config["num_classes"])

    print('Loading weights into state dict...')
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    model_dict = model.state_dict()
    pretrained_dict = torch.load("model_data/ssd_weights.pth", map_location=device)
    pretrained_dict = {k: v for k, v in pretrained_dict.items() if np.shape(model_dict[k]) ==  np.shape(v)}
    model_dict.update(pretrained_dict)
    model.load_state_dict(model_dict)
    print('Finished!')

    net = model.train()
    if Cuda:
        net = torch.nn.DataParallel(model)
        cudnn.benchmark = True
        net = net.cuda()

    annotation_path = '2007_train.txt'
    with open(annotation_path) as f:
        lines = f.readlines()
    np.random.seed(10101)
    np.random.shuffle(lines)
    np.random.seed(None)
    num_train = len(lines)

    if Use_Data_Loader:
        train_dataset = SSDDataset(lines[:num_train], (Config["min_dim"], Config["min_dim"]))
        gen = DataLoader(train_dataset, batch_size=Batch_size, num_workers=8, pin_memory=True,
                                drop_last=True, collate_fn=ssd_dataset_collate)
    else:
        gen = Generator(Batch_size, lines,
                        (Config["min_dim"], Config["min_dim"]), Config["num_classes"]).generate()

    criterion = MultiBoxLoss(Config['num_classes'], 0.5, True, 0, True, 3, 0.5,
                             False, Cuda)
    epoch_size = num_train // Batch_size

    if True:
        # ------------------------------------#
        #   冻结一定部分训练
        # ------------------------------------#
        for param in model.vgg.parameters():
            param.requires_grad = False

        optimizer = optim.Adam(net.parameters(), lr=lr)
        lr_scheduler = optim.lr_scheduler.StepLR(optimizer,step_size=1,gamma=0.95)
        for epoch in range(Start_iter,Freeze_epoch):
            with tqdm(total=epoch_size,desc=f'Epoch {epoch + 1}/{Freeze_epoch}',postfix=dict,mininterval=0.3) as pbar:
                loc_loss = 0
                conf_loss = 0
                for iteration, batch in enumerate(gen):
                    if iteration >= epoch_size:
                        break
                    images, targets = batch[0], batch[1]
                    with torch.no_grad():
                        if Cuda:
                            images = Variable(torch.from_numpy(images).type(torch.FloatTensor)).cuda()
                            targets = [Variable(torch.from_numpy(ann).type(torch.FloatTensor)).cuda() for ann in targets]
                        else:
                            images = Variable(torch.from_numpy(images).type(torch.FloatTensor))
                            targets = [Variable(torch.from_numpy(ann).type(torch.FloatTensor)) for ann in targets]
                    # 前向传播
                    out = net(images)
                    # 清零梯度
                    optimizer.zero_grad()
                    # 计算loss
                    loss_l, loss_c = criterion(out, targets)
                    loss = loss_l + loss_c
                    # 反向传播
                    loss.backward()
                    optimizer.step()
                    # 加上
                    loc_loss += loss_l.item()
                    conf_loss += loss_c.item()

                    pbar.set_postfix(**{'loc_loss'  : loc_loss / (iteration + 1), 
                                        'conf_loss' : conf_loss / (iteration + 1),
                                        'lr'        : get_lr(optimizer)})
                    pbar.update(1)
                        
            lr_scheduler.step()
            print('Saving state, iter:', str(epoch+1))
            torch.save(model.state_dict(), 'logs/Epoch%d-Loc%.4f-Conf%.4f.pth'%((epoch+1),loc_loss/(iteration+1),conf_loss/(iteration+1)))

    if True:
        # ------------------------------------#
        #   全部解冻训练
        # ------------------------------------#
        for param in model.vgg.parameters():
            param.requires_grad = True

        optimizer = optim.Adam(net.parameters(), lr=freeze_lr)
        lr_scheduler = optim.lr_scheduler.StepLR(optimizer,step_size=1,gamma=0.95)
        for epoch in range(Freeze_epoch,Epoch):
            with tqdm(total=epoch_size,desc=f'Epoch {epoch + 1}/{Freeze_epoch}',postfix=dict,mininterval=0.3) as pbar:
                loc_loss = 0
                conf_loss = 0
                for iteration, batch in enumerate(gen):
                    if iteration >= epoch_size:
                        break
                    images, targets = batch[0], batch[1]
                    with torch.no_grad():
                        if Cuda:
                            images = Variable(torch.from_numpy(images).type(torch.FloatTensor)).cuda()
                            targets = [Variable(torch.from_numpy(ann).type(torch.FloatTensor)).cuda() for ann in targets]
                        else:
                            images = Variable(torch.from_numpy(images).type(torch.FloatTensor))
                            targets = [Variable(torch.from_numpy(ann).type(torch.FloatTensor)) for ann in targets]
                    # 前向传播
                    out = net(images)
                    # 清零梯度
                    optimizer.zero_grad()
                    # 计算loss
                    loss_l, loss_c = criterion(out, targets)
                    loss = loss_l + loss_c
                    # 反向传播
                    loss.backward()
                    optimizer.step()
                    # 加上
                    loc_loss += loss_l.item()
                    conf_loss += loss_c.item()

                    pbar.set_postfix(**{'loc_loss'  : loc_loss / (iteration + 1), 
                                        'conf_loss' : conf_loss / (iteration + 1),
                                        'lr'        : get_lr(optimizer)})
                    pbar.update(1)

            lr_scheduler.step()
            print('Saving state, iter:', str(epoch+1))
            torch.save(model.state_dict(), 'logs/Epoch%d-Loc%.4f-Conf%.4f.pth'%((epoch+1),loc_loss/(iteration+1),conf_loss/(iteration+1)))