dense_model.py

import re
from typing import Any, List, Tuple
from collections import OrderedDict
from my_dataset import MyDataSet
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as cp
from torch import Tensor
import os
import math
import torch.optim as optim
import numpy as np
import pandas as pd
from torchvision import transforms
import torch.optim.lr_scheduler as lr_scheduler
from utils import train_one_epoch, evaluate, cal_m


class _DenseLayer(nn.Module):
    def __init__(self,
                 input_c: int,
                 growth_rate: int,
                 bn_size: int,
                 drop_rate: float,
                 memory_efficient: bool = False):
        super(_DenseLayer, self).__init__()
        self.add_module("norm1", nn.BatchNorm2d(input_c))
        self.add_module("relu1", nn.ReLU(inplace=True))
        self.add_module("conv1", nn.Conv2d(in_channels=input_c,
                                           out_channels=bn_size * growth_rate,
                                           kernel_size=1,
                                           stride=1,
                                           bias=False))
        self.add_module("norm2", nn.BatchNorm2d(bn_size * growth_rate))
        self.add_module("relu2", nn.ReLU(inplace=True))
        self.add_module("conv2", nn.Conv2d(bn_size * growth_rate,
                                           growth_rate,
                                           kernel_size=3,
                                           stride=1,
                                           padding=1,
                                           bias=False))
        self.drop_rate = drop_rate
        self.memory_efficient = memory_efficient

    def bn_function(self, inputs: List[Tensor]) -> Tensor:
        concat_features = torch.cat(inputs, 1)
        bottleneck_output = self.conv1(self.relu1(self.norm1(concat_features)))
        return bottleneck_output

    @staticmethod
    def any_requires_grad(inputs: List[Tensor]) -> bool:
        for tensor in inputs:
            if tensor.requires_grad:
                return True

        return False

    @torch.jit.unused
    def call_checkpoint_bottleneck(self, inputs: List[Tensor]) -> Tensor:
        def closure(*inp):
            return self.bn_function(inp)

        return cp.checkpoint(closure, *inputs)

    def forward(self, inputs: Tensor) -> Tensor:
        if isinstance(inputs, Tensor):
            prev_features = [inputs]
        else:
            prev_features = inputs
        if self.memory_efficient and self.any_requires_grad(prev_features):
            if torch.jit.is_scripting():
                raise Exception("memory efficient not supported in JIT")
            bottleneck_output = self.call_checkpoint_bottleneck(prev_features)
        else:
            bottleneck_output = self.bn_function(prev_features)
        new_features = self.conv2(self.relu2(self.norm2(bottleneck_output)))
        if self.drop_rate > 0:
            new_features = F.dropout(new_features,
                                     p=self.drop_rate,
                                     training=self.training)
        return new_features


class _DenseBlock(nn.ModuleDict):
    _version = 2

    def __init__(self,
                 num_layers: int,
                 input_c: int,
                 bn_size: int,
                 growth_rate: int,
                 drop_rate: float,
                 memory_efficient: bool = False):
        super(_DenseBlock, self).__init__()
        for i in range(num_layers):
            layer = _DenseLayer(input_c + i * growth_rate,
                                growth_rate=growth_rate,
                                bn_size=bn_size,
                                drop_rate=drop_rate,
                                memory_efficient=memory_efficient)
            self.add_module("denselayer%d" % (i + 1), layer)

    def forward(self, init_features: Tensor) -> Tensor:
        features = [init_features]
        for name, layer in self.items():
            new_features = layer(features)
            features.append(new_features)
        return torch.cat(features, 1)


class _Transition(nn.Sequential):
    def __init__(self,
                 input_c: int,
                 output_c: int):
        super(_Transition, self).__init__()
        self.add_module("norm", nn.BatchNorm2d(input_c))
        self.add_module("relu", nn.ReLU(inplace=True))
        self.add_module("conv", nn.Conv2d(input_c,
                                          output_c,
                                          kernel_size=1,
                                          stride=1,
                                          bias=False))
        self.add_module("pool", nn.AvgPool2d(kernel_size=2, stride=2))


class DenseNet(nn.Module):

    def __init__(self,
                 growth_rate: int = 32,
                 block_config: Tuple[int, int, int, int] = (6, 12, 24, 16),
                 num_init_features: int = 64,
                 bn_size: int = 4,
                 drop_rate: float = 0,
                 num_classes: int = 1000,
                 memory_efficient: bool = False):
        super(DenseNet, self).__init__()

        # first conv+bn+relu+pool
        self.features = nn.Sequential(OrderedDict([
            ("conv0", nn.Conv2d(3, num_init_features, kernel_size=7, stride=2, padding=3, bias=False)),
            ("norm0", nn.BatchNorm2d(num_init_features)),
            ("relu0", nn.ReLU(inplace=True)),
            ("pool0", nn.MaxPool2d(kernel_size=3, stride=2, padding=1)),
        ]))

        num_features = num_init_features
        for i, num_layers in enumerate(block_config):
            block = _DenseBlock(num_layers=num_layers,
                                input_c=num_features,
                                bn_size=bn_size,
                                growth_rate=growth_rate,
                                drop_rate=drop_rate,
                                memory_efficient=memory_efficient)
            self.features.add_module("denseblock%d" % (i + 1), block)
            num_features = num_features + num_layers * growth_rate

            if i != len(block_config) - 1:
                trans = _Transition(input_c=num_features,
                                    output_c=num_features // 2)
                self.features.add_module("transition%d" % (i + 1), trans)
                num_features = num_features // 2

        self.features.add_module("norm5", nn.BatchNorm2d(num_features))

        self.classifier = nn.Linear(num_features, num_classes)

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight)
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)
            elif isinstance(m, nn.Linear):
                nn.init.constant_(m.bias, 0)

    def forward(self, x: Tensor) -> Tensor:
        features = self.features(x)
        out = F.relu(features, inplace=True)
        out = F.adaptive_avg_pool2d(out, (1, 1))
        out = torch.flatten(out, 1)
        out = self.classifier(out)
        return out


def densenet121(**kwargs: Any) -> DenseNet:
    return DenseNet(growth_rate=32,
                    block_config=(6, 12, 24, 16),
                    num_init_features=64,
                    **kwargs)


def densenet169(**kwargs: Any) -> DenseNet:
    return DenseNet(growth_rate=32,
                    block_config=(6, 12, 32, 32),
                    num_init_features=64,
                    **kwargs)


def densenet201(**kwargs: Any) -> DenseNet:
    return DenseNet(growth_rate=32,
                    block_config=(6, 12, 48, 32),
                    num_init_features=64,
                    **kwargs)


def densenet161(**kwargs: Any) -> DenseNet:
    return DenseNet(growth_rate=48,
                    block_config=(6, 12, 36, 24),
                    num_init_features=96,
                    **kwargs)


def load_state_dict(model: nn.Module, weights_path: str) -> None:
    pattern = re.compile(
        r'^(.*denselayer\d+\.(?:norm|relu|conv))\.((?:[12])\.(?:weight|bias|running_mean|running_var))$')
    state_dict = torch.load(weights_path)
    num_classes = model.classifier.out_features
    load_fc = num_classes == 1000

    for key in list(state_dict.keys()):
        if load_fc is False:
            if "classifier" in key:
                del state_dict[key]

        res = pattern.match(key)
        if res:
            new_key = res.group(1) + res.group(2)
            state_dict[new_key] = state_dict[key]
            del state_dict[key]
    model.load_state_dict(state_dict, strict=load_fc)
    print("successfully load pretrain-weights.")


def train(train_images_path, train_images_label, val_images_path=None, val_images_label=None, classes=2, epochs=20,
          batch_size=32,
          lr=0.001, lrf=0.01, weight='./init/densenet121.pth', freeze=False, device='cuda:0', val=True, data=1, init=True):
    device = torch.device(device if torch.cuda.is_available() else "cpu")
    # tb_writer = SummaryWriter()
    if val_images_label is None:
        val_images_label = []
    if val_images_path is None:
        val_images_path = []
    data_transform = {
        "train": transforms.Compose([transforms.RandomResizedCrop(224),
                                     transforms.RandomHorizontalFlip(),
                                     transforms.ToTensor(),
                                     transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])]),
        "val": transforms.Compose([transforms.Resize(256),
                                   transforms.CenterCrop(224),
                                   transforms.ToTensor(),
                                   transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])}

    train_dataset = MyDataSet(images_path=train_images_path,
                              images_class=train_images_label,
                              transform=data_transform["train"])
    if val:
        val_dataset = MyDataSet(images_path=val_images_path,
                                images_class=val_images_label,
                                transform=data_transform["val"])

    nw = min([os.cpu_count(), batch_size if batch_size > 1 else 0, 8])
    print('Using {} dataloader workers every process'.format(nw))
    train_loader = torch.utils.data.DataLoader(train_dataset,
                                               batch_size=batch_size,
                                               shuffle=True,
                                               pin_memory=True,
                                               num_workers=nw,
                                               collate_fn=train_dataset.collate_fn)
    if val:
        val_loader = torch.utils.data.DataLoader(val_dataset,
                                                 batch_size=batch_size,
                                                 shuffle=False,
                                                 pin_memory=True,
                                                 num_workers=nw,
                                                 collate_fn=val_dataset.collate_fn)

    model = densenet121(num_classes=classes).to(device)
    if init:
        if os.path.exists(weight):
            load_state_dict(model, weight)
        else:
            raise FileNotFoundError("not found weights file: {}".format(weight))

    if freeze:
        for name, para in model.named_parameters():
            if "classifier" not in name:
                para.requires_grad_(False)

    pg = [p for p in model.parameters() if p.requires_grad]
    optimizer = optim.SGD(pg, lr=lr, momentum=0.9, weight_decay=1E-4, nesterov=True)
    lf = lambda x: ((1 + math.cos(x * math.pi / epochs)) / 2) * (1 - lrf) + lrf
    scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)
    best_acc = 0.0
    best_epoch = 0
    for epoch in range(epochs):
        train_loss, train_acc = train_one_epoch(model=model,
                                                optimizer=optimizer,
                                                data_loader=train_loader,
                                                device=device,
                                                epoch=epoch)
        scheduler.step()
        val_loss, val_acc = 0, 0
        if val:
            val_loss, val_acc = evaluate(model=model,
                                         data_loader=val_loader,
                                         device=device,
                                         epoch=epoch)

        f = open('./res_dir/train_res/data' + str(data) + '/densenet_res.txt', 'a')
        f.write('epoch: ' + str(epoch + 1) + '\n')
        f.write("train_loss: " + str(round(train_loss, 4)) + '\n')
        f.write("train_acc: " + str(round(train_acc, 4)) + '\n')
        f.write("val_loss: " + str(round(val_loss, 4)) + '\n')
        f.write("val_acc: " + str(round(val_acc, 4)) + '\n\n')
        f.close()
        if best_acc < train_acc:
            best_acc = train_acc
            best_epoch = epoch
            torch.save(model.state_dict(), './res_dir/weights/data' + str(data) + '/best_densenet_model.pth')
    f1 = open('./res_dir/best_res/best.txt', 'a')
    f1.write('dense ' + 'epoch ' + str(best_epoch) + '  ' + str(best_acc) + '\n')


def predict(images_path, images_label, num_class=2, batch_size=8, data=1):
    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
    data_transform = transforms.Compose(
        [transforms.Resize(256),
         transforms.CenterCrop(224),
         transforms.ToTensor(),
         transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])
    dataset = MyDataSet(images_path=images_path,
                        images_class=images_label,
                        transform=data_transform)
    nw = min([os.cpu_count(), batch_size if batch_size > 1 else 0, 8])
    loader = torch.utils.data.DataLoader(dataset,
                                         batch_size=batch_size,
                                         shuffle=False,
                                         pin_memory=True,
                                         num_workers=nw,
                                         collate_fn=dataset.collate_fn)
    model = densenet121(num_classes=num_class).to(device)
    model_weight_path = './res_dir/weights/data' + str(data) + '/best_densenet_model.pth'
    model.load_state_dict(torch.load(model_weight_path, map_location=device))
    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
    score = cal_m(model=model, data_loader=loader, device=device, alo='dense', num=num_class)
    pd.DataFrame(np.array(score)).to_csv('./res_dir/label/data' + str(data) + '/dense' + '.csv', header=None, index=None)
    return score