EffNet图像分类--分类精子图片

凯哥英语视频

1.现有资源梳理

人为分成四类的精子视野视频几千个
四种分类截图实例
0

1

2

3

共四类，
0–精子数量合理，交叉少，运动分析算法能识别
1–精子数量很多，交叉多，堆积多，运动分析算法无法识别
2–精子数量合理，交叉多，运动分析算法较难识别
3–精子数量合理，精子多层，运动分析算法较难识别

操作环境 win10-64位
代码语言 Python 3.6
运行环境 torch 显存尽量多吧，小的话训练的比较慢

2.实现方案规划

数据集准备

1. 视频转图片–50帧取一张
2. 生成对应的 train_data.txt test_data.txt 等文件

参数调整

1. 基础模型选择b0-b8（暂定的b8）
2. batch-size（我的图片比较大1536X1024所以就设定了4）
3. data，train_data.txt等文件的位置设定

训练及验证

1. 准备好数据等，运行main_PM_yk.py训练，等吧。。。
2. 准备好数据等，运行peidect_100.py测试，等吧。。。

没了。。。

3.代码实现（代码有点多，这里只展示主要的）

要改的地方我都标注了
训练文件main_PM_yk.py

"""
Evaluate on ImageNet. Note that at the moment, training is not implemented (I am working on it).
that being said, evaluation is working.
"""

import argparse
import os
import random
import shutil
import time
import warnings
import PIL
import pathlib
import collections
import utils
import numpy as np
from tensorboardX import SummaryWriter
is_tensorboard_available = True


import torch
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.distributed as dist
import torch.optim
import torch.multiprocessing as mp
import torch.utils.data
import torch.utils.data.distributed
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import torchvision.models as models
from sam import SAM
from efficientnet_pytorch import EfficientNet
from dataset import MyDataset
# from dataset_loc import MyDataset

parser = argparse.ArgumentParser(description='PyTorch ImageNet Training')
parser.add_argument('--data', metavar='DIR',default='./Original_video_file/',type=str,
                    help='path to dataset')
###   模型 选择b0-b8
parser.add_argument('-a', '--arch', metavar='ARCH', default='efficientnet-b8',
                    help='model architecture (default: resnet18)')
parser.add_argument('-j', '--workers', default=40, type=int, metavar='N',
                    help='number of data loading workers (default: 4)')
parser.add_argument('--epochs', default=100, type=int, metavar='N',
                    help='number of total epochs to run')
parser.add_argument('--start_epoch', default=0, type=int, metavar='N',
                    help='manual epoch number (useful on restarts)')
###   每轮数据训练数                   
parser.add_argument('-b', '--batch-size', default=4, type=int,
                    metavar='N',
                    help='mini-batch size (default: 256), this is the total '
                         'batch size of all GPUs on the current node when '
                         'using Data Parallel or Distributed Data Parallel')
parser.add_argument('--lr', '--learning_rate', default=0.01, type=float,
                    metavar='LR', help='initial learning rate', dest='lr')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
                    help='momentum')
parser.add_argument('--wd', '--weight-decay', default=1e-4, type=float,
                    metavar='W', help='weight decay (default: 1e-4)',
                    dest='weight_decay')
parser.add_argument('-p', '--print-freq', default=10, type=int,
                    metavar='N', help='print frequency (default: 10)')
parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_true',
                    default=False,help='evaluate model on validation set')
parser.add_argument('--pretrained', dest='pretrained', action='store_true',
                    default=True,help='use pre-trained model')
parser.add_argument('--rank', default=0, type=int,
                    help='node rank for distributed training')
parser.add_argument('--dist-backend', default='nccl', type=str,
                    help='distributed backend')
parser.add_argument('--seed', default=None, type=int,
                    help='seed for initializing training. ')
parser.add_argument('--gpu', default=None, type=int,
                    help='GPU id to use.')
parser.add_argument('--image_size', default=224, type=int,
                    help='image size')
parser.add_argument('--advprop', default=True, action='store_true',
                    help='use advprop or not')
parser.add_argument('--multiprocessing-distributed', action='store_true',
                    default=False,
                    help='Use multi-processing distributed training to launch '
                         'N processes per node, which has N GPUs. This is the '
                         'fastest way to use PyTorch for either single node or '
                         'multi node data parallel training')
parser.add_argument("--rho", default=0.05, type=int, help="Rho parameter for SAM.")
use_tensorboard = True

outdir = './result/E8_P_0111_SAM_smooth0.1_PM4lable'
filename = './result/E8_P_0111_SAM_smooth0.1_PM4lable'
outdir = pathlib.Path(outdir)
outdir.mkdir(exist_ok=True, parents=True)
best_acc1 = 0
if use_tensorboard:
    writer = SummaryWriter(os.path.join(outdir, 'writer'))
    print('use tensorboard file is writer')

def main():
    args = parser.parse_args()

    if args.seed is not None:
        random.seed(args.seed)
        torch.manual_seed(args.seed)
        cudnn.deterministic = True
        warnings.warn('You have chosen to seed training. '
                      'This will turn on the CUDNN deterministic setting, '
                      'which can slow down your training considerably! '
                      'You may see unexpected behavior when restarting '
                      'from checkpoints.')

    if args.gpu is not None:
        warnings.warn('You have chosen a specific GPU. This will completely '
                      'disable data parallelism.')

    ngpus_per_node = torch.cuda.device_count()
    # Simply call main_worker function
    main_worker(args.gpu, ngpus_per_node, args)


def main_worker(gpu, ngpus_per_node, args):
    global best_acc1
    args.gpu = gpu
    
    # create model
    if 'efficientnet' in args.arch:  # NEW
        if args.pretrained:
            if args.start_epoch != 0:
                model = EfficientNet.from_name('efficientnet-b8', override_params={'num_classes': 3})
                # print(model)
                eff_model = torch.load('result/E8_P_0520_SGD0.01_1_4.5_25.5/model_best9.pth')['state_dict'].items()
                model_wri = {}
                for k,v in eff_model:
                    name = k[7:]
                    if name not in ['_fc.weight','_fc.bias']:
                        model_wri[name]=v
                fc_w = torch.empty(3,2816)
                fc_b = torch.empty(3)
                model_wri['_fc.weight'] = nn.init.kaiming_normal_(fc_w)
                model_wri['_fc.bias'] = nn.init.constant_(fc_b, 0)
                
                model.load_state_dict(model_wri)
                print("=> using pre-trained model '{}' and learn".format(args.arch))
                
            ####     模型分类数
            else:
                model = EfficientNet.from_pretrained(args.arch,num_classes=4,advprop=args.advprop)
                print("=> using pre-trained model '{}'".format(args.arch))
        else:
            print("=> creating model '{}'".format(args.arch,num_classes=2))
            model = EfficientNet.from_name(args.arch,override_params={'num_classes': 2})

    else:
        if args.pretrained:
            print("=> using pre-trained model '{}'".format(args.arch))
            model = models.__dict__[args.arch](pretrained=True)
        else:
            print("=> creating model '{}'".format(args.arch))
            model = models.__dict__[args.arch]()

 
    if args.arch.startswith('alexnet') or args.arch.startswith('vgg'):
        model.features = torch.nn.DataParallel(model.features)
        model.cuda()
    else:
        model = torch.nn.DataParallel(model).cuda()

    # define loss function (criterion) and optimizer
    # weight = torch.from_numpy(np.array([1,2,2])).float()
    criterion = None

    # optimizer = torch.optim.SGD(model.parameters(), args.lr,
                                # momentum=args.momentum,
                                # weight_decay=args.weight_decay)
    base_optimizer = torch.optim.SGD  # define an optimizer for the "sharpness-aware" update
    optimizer = SAM(model.parameters(), base_optimizer, rho=args.rho, lr=args.lr, momentum=args.momentum, weight_decay=args.weight_decay)
    #optimizer = torch.optim.Adam(model.parameters(), lr=args.lr, weight_decay=args.weight_decay)

    cudnn.benchmark = True

    # Data loading code
    # traindir = os.path.join(args.data, 'train')
    # valdir = os.path.join(args.data, 'val')
    if args.advprop:
        normalize = transforms.Lambda(lambda img: img * 2.0 - 1.0)
    else:
        normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                         std=[0.229, 0.224, 0.225])
    ####     训练TXT                                 
    train_dataset = MyDataset(
        args.data,
        'txt/train_data.txt',
        transform=transforms.Compose([
            transforms.RandomOrder([
            transforms.RandomHorizontalFlip(p=0.5),
            transforms.RandomVerticalFlip(p=0.5),
            transforms.RandomRotation(20),
            transforms.ColorJitter(brightness=0.1, contrast=0.2) 
            ]),
            transforms.Resize((EfficientNet.get_image_size(args.arch),EfficientNet.get_image_size(args.arch)), interpolation=PIL.Image.BICUBIC),
            transforms.ToTensor(),
            normalize,
        ]),
        ifp = True)
 
    train_sampler = None

    train_loader = torch.utils.data.DataLoader(
        dataset=train_dataset, batch_size=args.batch_size, shuffle=(train_sampler is None),
        num_workers=args.workers, pin_memory=True, sampler=train_sampler)

    if 'efficientnet' in args.arch:
        image_size = EfficientNet.get_image_size(args.arch)
        val_transforms = transforms.Compose([
            transforms.Resize((image_size,image_size), interpolation=PIL.Image.BICUBIC),
            transforms.ToTensor(),
            normalize,
        ])
        print('Using image size', image_size)
    ####     验证TXT  因为懒···所以测试集跟训练集一样···可以自己分
    val_loader = torch.utils.data.DataLoader(
        dataset=MyDataset(args.data, 'txt/train_data.txt', transform=val_transforms,ifp=True),
        batch_size=args.batch_size, shuffle=False,
        num_workers=args.workers, pin_memory=True)
    
    if args.evaluate:
        res = validate(val_loader, model, criterion, args)
        with open('res.txt', 'w') as f:
            print(res, file=f)
        return
    
    epoch_logs = []
    test_loss_best = 10000
    test_acc_best = 0
    
    for epoch in range(args.start_epoch, args.epochs):
        adjust_learning_rate(optimizer, epoch, args)

        # train for one epoch
        train_log = train(train_loader, model, criterion, optimizer, epoch, args)

        # evaluate on validation set
        acc1,test_log,test_loss = validate(val_loader, model, criterion, args, epoch)

        # remember best acc@1 and save checkpoint
        is_best = acc1 >= best_acc1
        best_acc1 = max(acc1, best_acc1)

        if not args.multiprocessing_distributed or (args.multiprocessing_distributed
                and args.rank % ngpus_per_node == 0):
            save_checkpoint({
                'epoch': epoch + 1,
                'arch': args.arch,
                'state_dict': model.state_dict(),
                'best_acc1': best_acc1,
                # 'optimizer' : optimizer.state_dict(),
            }, is_best, filename=filename)
            
        epoch_log = train_log.copy()
        epoch_log.update(test_log)
        epoch_logs.append(epoch_log)
        # print(epoch_logs)
        utils.save_epoch_logs(epoch_logs, outdir)
        
        if test_acc_best < acc1:
            test_acc_best = acc1
            num_test = 0
        else:
            num_test += 1
            
        if num_test >= 25:
            print('训练结束验证集acc停止上升')
            break
        
        # if test_loss_best > test_loss:
            # test_loss_best = test_loss
            # num_test = 0
        # else:
            # num_test += 1
        # if num_test >= 25:
            # print('训练结束验证集Loss停止下降')
            # break
import torch.nn.functional as F
def smooth_crossentropy(pred, gold, smoothing=0.1):
    n_class = pred.size(1)
    one_hot = torch.full_like(pred, fill_value=smoothing / (n_class - 1))
    one_hot.scatter_(dim=1, index=gold.unsqueeze(1), value=1.0 - smoothing)
    log_prob = F.log_softmax(pred, dim=1)
    return F.kl_div(input=log_prob, target=one_hot, reduction='none').sum(-1)   

def train(train_loader, model, criterion, optimizer, epoch, args):
    batch_time = AverageMeter('Time', ':6.3f')
    data_time = AverageMeter('Data', ':6.3f')
    losses = AverageMeter('Loss', ':.4e')
    top1 = AverageMeter('Acc@1', ':6.2f')
    top2 = AverageMeter('Acc@2', ':6.2f')
    progress = ProgressMeter(len(train_loader), batch_time, data_time, losses, top1,
                             top2, prefix="Epoch: [{}]".format(epoch))

    # switch to train mode
    model.train()

    end = time.time()
    for i, (images, target) in enumerate(train_loader):
        # measure data loading time
        data_time.update(time.time() - end)

        if args.gpu is not None:
            images = images.cuda(args.gpu, non_blocking=True)
        target = target.cuda(args.gpu, non_blocking=True)

        # compute output
        output = model(images)
        loss = smooth_crossentropy(output, target)
        loss.mean().backward()
        optimizer.first_step(zero_grad=True)
        smooth_crossentropy(model(images), target).mean().backward()
        optimizer.second_step(zero_grad=True)
        
        # measure accuracy and record loss
        acc1, acc2 = accuracy(output, target, topk=(1, 2))
        # print(loss, images.size(0))
        losses.update(loss.mean(), images.size(0))
        top1.update(acc1[0], images.size(0))
        top2.update(acc2[0], images.size(0))

        # compute gradient and do SGD step
        # optimizer.zero_grad()
        
        # measure elapsed time
        batch_time.update(time.time() - end)
        end = time.time()

        if i % args.print_freq == 0:
            progress.print(i)
        
    if use_tensorboard:
        writer.add_scalar('Train/Loss', losses.avg, epoch)
        writer.add_scalar('Train/Accuracy', top1.avg, epoch)
        writer.add_scalar('Train/Time', batch_time.avg, epoch)

    train_log = collections.OrderedDict({
        'epoch':
        epoch,
        'train':
        collections.OrderedDict({
            'loss': str(losses.avg),
            'accuracy': float('{top1.avg:.3f}'.format(top1=top1)),
            'time': batch_time.avg,
        }),
    })
    return train_log


def validate(val_loader, model, criterion, args, epoch):
    batch_time = AverageMeter('Time', ':6.3f')
    losses = AverageMeter('Loss', ':.4e')
    top1 = AverageMeter('Acc@1', ':6.2f')
    top2 = AverageMeter('Acc@2', ':6.2f')
    progress = ProgressMeter(len(val_loader), batch_time, losses, top1, top2,
                             prefix='Test: ')

    # switch to evaluate mode
    model.eval()

    with torch.no_grad():
        end = time.time()
        for i, (images, target) in enumerate(val_loader):
            if args.gpu is not None:
                images = images.cuda(args.gpu, non_blocking=True)
            target = target.cuda(args.gpu, non_blocking=True)

            # compute output
            output = model(images)
            loss = smooth_crossentropy(output, target)

            # measure accuracy and record loss
            acc1, acc2 = accuracy(output, target, topk=(1, 2))
            losses.update(loss.mean(), images.size(0))
            top1.update(acc1[0], images.size(0))
            top2.update(acc2[0], images.size(0))

            # measure elapsed time
            batch_time.update(time.time() - end)
            end = time.time()

            if i % args.print_freq == 0:
                progress.print(i)

        # TODO: this should also be done with the ProgressMeter
        print(' * Acc@1 {top1.avg:.3f} Acc@2 {top2.avg:.3f}'
              .format(top1=top1, top2=top2))
        
    if use_tensorboard:
        writer.add_scalar('Test/Loss', losses.avg, epoch)
        writer.add_scalar('Test/Accuracy', top1.avg, epoch)
        writer.add_scalar('Test/Time', batch_time.avg, epoch)

    test_log = collections.OrderedDict({
        'epoch':
        epoch,
        'train':
        collections.OrderedDict({
            'loss': str(losses.avg),
            'accuracy': float('{top1.avg:.3f}'.format(top1=top1)),
            'time': batch_time.avg,
        }),
    })

    return top1.avg,test_log,losses.avg


def save_checkpoint(state, is_best, filename='./result/'):
    torch.save(state, filename+'/checkpoint.pth')
    if is_best:
        shutil.copyfile(filename+'/checkpoint.pth', filename+'/model_best'+str(state['epoch'])+'.pth')


class AverageMeter(object):
    """Computes and stores the average and current value"""
    def __init__(self, name, fmt=':f'):
        self.name = name
        self.fmt = fmt
        self.reset()

    def reset(self):
        self.val = 0
        self.avg = 0
        self.sum = 0
        self.count = 0

    def update(self, val, n=1):
        self.val = val
        self.sum += val * n
        self.count += n
        self.avg = self.sum / self.count

    def __str__(self):
        fmtstr = '{name} {val' + self.fmt + '} ({avg' + self.fmt + '})'
        return fmtstr.format(**self.__dict__)


class ProgressMeter(object):
    def __init__(self, num_batches, *meters, prefix=""):
        self.batch_fmtstr = self._get_batch_fmtstr(num_batches)
        self.meters = meters
        self.prefix = prefix

    def print(self, batch):
        entries = [self.prefix + self.batch_fmtstr.format(batch)]
        entries += [str(meter) for meter in self.meters]
        print('\t'.join(entries))

    def _get_batch_fmtstr(self, num_batches):
        num_digits = len(str(num_batches // 1))
        fmt = '{:' + str(num_digits) + 'd}'
        return '[' + fmt + '/' + fmt.format(num_batches) + ']'


def adjust_learning_rate(optimizer, epoch, args):
    """Sets the learning rate to the initial LR decayed by 10 every 30 epochs"""
    lr = args.lr * (0.1 ** (epoch // 20))
    for param_group in optimizer.param_groups:
        param_group['lr'] = lr


def accuracy(output, target, topk=(1,)):
    """Computes the accuracy over the k top predictions for the specified values of k"""
    with torch.no_grad():
        maxk = max(topk)
        batch_size = target.size(0)

        _, pred = output.topk(maxk, 1, True, True)
        pred = pred.t()
        correct = pred.eq(target.view(1, -1).expand_as(pred))

        res = []
        for k in topk:
            correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
            res.append(correct_k.mul_(100.0 / batch_size))
        return res


if __name__ == '__main__':
    main()

验证文件peidect_100.py

"""
Evaluate on ImageNet. Note that at the moment, training is not implemented (I am working on it).
that being said, evaluation is working.
"""

import argparse
import os
import random
import shutil
import time
import warnings
import PIL
import pathlib
import collections
import pandas as pd
import os

import torch
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.distributed as dist
import torch.optim
import torch.multiprocessing as mp
import torch.utils.data
import torch.utils.data.distributed
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import torchvision.models as models

from efficientnet_pytorch import EfficientNet
from dataset import MyDataset

parser = argparse.ArgumentParser(description='PyTorch ImageNet Training')
parser.add_argument('--data', metavar='DIR', default='./Original_video_file/', type=str,
                    help='path to dataset')
parser.add_argument('-a', '--arch', metavar='ARCH', default='efficientnet-b8',
                    help='model architecture (default: resnet18)')
parser.add_argument('-j', '--workers', default=15, type=int, metavar='N',
                    help='number of data loading workers (default: 4)')
parser.add_argument('--epochs', default=100, type=int, metavar='N',
                    help='number of total epochs to run')
parser.add_argument('--start-epoch', default=0, type=int, metavar='N',
                    help='manual epoch number (useful on restarts)')
parser.add_argument('-b', '--batch-size', default=1, type=int,
                    metavar='N',
                    help='mini-batch size (default: 256), this is the total '
                         'batch size of all GPUs on the current node when '
                         'using Data Parallel or Distributed Data Parallel')
parser.add_argument('--lr', '--learning-rate', default=0.1, type=float,
                    metavar='LR', help='initial learning rate', dest='lr')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
                    help='momentum')
parser.add_argument('--wd', '--weight-decay', default=1e-4, type=float,
                    metavar='W', help='weight decay (default: 1e-4)',
                    dest='weight_decay')
parser.add_argument('-p', '--print-freq', default=10, type=int,
                    metavar='N', help='print frequency (default: 10)')
parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_true',
                    default=True, help='evaluate model on validation set')
parser.add_argument('--pretrained', dest='pretrained', action='store_true',
                    default=True, help='use pre-trained model')
parser.add_argument('--world-size', default=0, type=int,
                    help='number of nodes for distributed training')
parser.add_argument('--rank', default=0, type=int,
                    help='node rank for distributed training')
parser.add_argument('--dist-url', default='tcp://224.66.41.62:23456', type=str,
                    help='url used to set up distributed training')
parser.add_argument('--gpu', default=None, type=int,
                    help='GPU id to use.')
parser.add_argument('--image_size', default=456, type=int,
                    help='image size')
parser.add_argument('--advprop', default=True, action='store_true',
                    help='use advprop or not')
parser.add_argument('--multiprocessing-distributed', action='store_true',
                    default=False,
                    help='Use multi-processing distributed training to launch '
                         'N processes per node, which has N GPUs. This is the '
                         'fastest way to use PyTorch for either single node or '
                         'multi node data parallel training')
with open('datas/PM_val.txt', 'r') as f:
    lables = f.read()
    lables = lables.split('\n')
    lable = [a.split()[0].split('/')[-1] for a in lables]
best_acc1 = 0


def main():
    args = parser.parse_args()
    if args.dist_url == "env://" and args.world_size == -1:
        args.world_size = int(os.environ["WORLD_SIZE"])

    main_worker(args.gpu, args)


def main_worker(gpu, args):
    global best_acc1
    args.gpu = gpu

    # create model
    if 'efficientnet' in args.arch:  # NEW

        if args.pretrained:
            model = EfficientNet.from_name('efficientnet-b8', override_params={'num_classes': 4})
            eff_model = torch.load('./result/E8_P_0111_SAM_smooth0.1_PM4lable/model_best20.pth')['state_dict'].items()
            model_wri = {}
            for k, v in eff_model:
                name = k[7:]
                model_wri[name] = v
            model.load_state_dict(model_wri)     
    model = torch.nn.DataParallel(model).cuda()
    cudnn.benchmark = True
    normalize = transforms.Lambda(lambda img: img * 2.0 - 1.0)
    image_size = EfficientNet.get_image_size(args.arch)
    val_transforms = transforms.Compose([
        transforms.Resize(image_size, interpolation=PIL.Image.BICUBIC),
        transforms.ToTensor(),
        normalize,
    ])
    print('Using image size', image_size)
    val_loader = torch.utils.data.DataLoader(
        dataset=MyDataset(args.data, 'PM_val.txt', transform=val_transforms,ifp=True),
        batch_size=args.batch_size, shuffle=False,
        num_workers=args.workers, pin_memory=True)
    # 预测
    if args.evaluate:
        res = validate(val_loader, model, args)
        with open('res.txt', 'w') as f:
            print(res, file=f)
        return


def validate(val_loader, model, args):
    # switch to evaluate mode
    model.eval()

    num = 0
    rig_0 = 0
    rig_1 = 0
    rig_2 = 0
    rig_3 = 0
    rig_4 = 0
    label_0 = 0
    label_1 = 0
    label_2 = 0
    label_3 = 0
    label_4 = 0
    lable_list = []
    pridect_list = []
    num_epoch = 1
    labels_list = []

    with torch.no_grad():
        for i, (images, target) in enumerate(val_loader):
            if args.gpu is not None:
                images = images.cuda(args.gpu, non_blocking=True)
            target = target.cuda(args.gpu, non_blocking=True)

            # compute output
            output = model(images)

            # measure accuracy and record loss
            _, preds = torch.max(output, dim=1)
            a = torch.nn.functional.softmax(output)
            
            preds = int(preds)

            target = int(target)
            num += 1
            if target == 0:
                label_0 +=1
                if preds == target:
                        rig_0+=1
            if target == 1:
                label_1 +=1
                if preds == target:
                        rig_1+=1
            if target == 2:
                label_2 +=1
                if preds == target:
                        rig_2+=1
            if target == 3:
                label_3 +=1
                if preds == target:
                        rig_3+=1

            
            
            labels_list.append(a)
            pridect_list.append(preds)
            lable_list.append(target)
            
            acc = (rig_0 + rig_1 + rig_2 + rig_3 + rig_4) / num
            print('{} pred:{},{}'.format(num_epoch, preds, target),
                  f'total:{num},rig_0:{rig_0},rig_1:{rig_1},rig_2:{rig_2},rig_3:{rig_3},rig_4:{rig_4},acc:{acc}')
            num_epoch += 1

        total_right = rig_0 + rig_1 + rig_2 + rig_3 + rig_4
        
        # print('total_right: {},total: {}, avg_acc:{},right_0: {}, label_0: {}, acc_0: {},right_1: {}, label_1: {}, acc_1: {},right_2: {}, label_2: {}, acc_2: {}'
        # .format(total_right,num,total_right/num,rig_0,label_0,rig_0/label_0,rig_1,label_1,rig_1/label_1,rig_2,label_2,rig_2/label_2))
        
        dataframe = pd.DataFrame(
            {'name': lable, 'pridect': pridect_list, 'lable':lable_list,'max':labels_list})
        dataframe.to_csv('PM_4.csv', index=False)

    return total_right


def accuracy(output, target, topk=(1,)):
    """Computes the accuracy over the k top predictions for the specified values of k"""
    with torch.no_grad():
        maxk = max(topk)
        batch_size = target.size(0)

        _, pred = output.topk(maxk, 1, True, True)
        pred = pred.t()
        correct = pred.eq(target.view(1, -1).expand_as(pred))

        res = []
        for k in topk:
            correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
            res.append(correct_k.mul_(100.0 / batch_size))
        return res


if __name__ == '__main__':
    main()
 

如果图片地址有问题，修改dataset.py

修改如下部分
img = cv2.imread(self.root + 'image/' + fn)

项目打包链接

项目有点大,我就放在百度云了
项目打包下载
提取密码 520y

结语

别的也没啥说的 , 如果觉得可以 , 希望一键三连支持一下 !

ok，那就这样吧~

欢迎各位大佬留言吐槽,也可以深入交流~