Pytorch: Wired results of model on testing dataset

I am using timm model to fine-tune resnet152 model on dataset contains 10 classes. The accuracy on validation dataset is good but when testing on nuisances classes model becomes performance becomes worst. Nuisances classes are very hard to detect and each classes consists of 10 classes. Like Nuisance class is weather and weather class consists of ten classes like aeroplane, car, truck and etc. I am unable to find the problem in model

Train Script

import argparse
import os
import random
import shutil
import ssl
import time
import warnings
from enum import Enum

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

ssl._create_default_https_context = ssl._create_unverified_context

import timm
from datasets import PoseData
from ssl_aug import rand_square_mask

model_names = sorted(name for name in models.__dict__
                     if name.islower() and not name.startswith("__")
                     and callable(models.__dict__[name]))

parser = argparse.ArgumentParser(description='ROBIN 3D Pose Estimation')
parser.add_argument('--data', help='path to train dataset', default='/home/lynk/Documents/ROBINv1.1-cls'
parser.add_argument('--val-data', help='path to validation dataset', default='/home/lynk/Documents'
parser.add_argument('-a', '--arch', metavar='ARCH', default='resnet152',
                    help='model architecture: ' +
                         ' | '.join(model_names) +
                         ' (default: resnet152)')
parser.add_argument('-j', '--workers', default=4, type=int, metavar='N',
                    help='number of data loading workers (default: 4)')
parser.add_argument('--epochs', default=200, 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=64, type=int,
                    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=1e-4, type=float,  # 1e-8
                    metavar='LR', help='initial learning rate', dest='lr')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
parser.add_argument('--grid-size', default=30, type=float, metavar='G',
                    help='grid size of the pose estimation')
parser.add_argument('--wd', '--weight-decay', default=0, type=float,
                    metavar='W', help='weight decay (default: 1e-4)',
parser.add_argument('-p', '--print-freq', default=10, type=int,
                    metavar='N', help='print frequency (default: 10)')
parser.add_argument('--resume', default='', type=str, metavar='PATH',
                    help='path to latest checkpoint (default: none)')
parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_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=-1, type=int,
                    help='number of nodes for distributed training')
parser.add_argument('--rank', default=-1, type=int,
                    help='node rank for distributed training')
parser.add_argument('--dist-url', default='tcp://', type=str,
                    help='url used to set up 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=0, type=int,
                    help='GPU id to use.')
parser.add_argument('--multiprocessing-distributed', action='store_true',
                    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')

best_acc1 = 0

args = parser.parse_args()

def main():
    if args.seed is not None:
        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.')

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

    args.distributed = args.world_size > 1 or args.multiprocessing_distributed

    ngpus_per_node = torch.cuda.device_count()
    if args.multiprocessing_distributed:
        # Since we have ngpus_per_node processes per node, the total world_size
        # needs to be adjusted accordingly
        args.world_size = ngpus_per_node * args.world_size
        # Use torch.multiprocessing.spawn to launch distributed processes: the
        # main_worker process function
        mp.spawn(main_worker, nprocs=ngpus_per_node, args=(ngpus_per_node, args))
        # 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

    if args.gpu is not None:
        print("Use GPU: {} for training".format(args.gpu))

    if args.distributed:
        if args.dist_url == "env://" and args.rank == -1:
            args.rank = int(os.environ["RANK"])
        if args.multiprocessing_distributed:
            # For multiprocessing distributed training, rank needs to be the
            # global rank among all the processes
            args.rank = args.rank * ngpus_per_node + gpu
        dist.init_process_group(backend=args.dist_backend, init_method=args.dist_url,
                                world_size=args.world_size, rank=args.rank)
    # create model

    # out_channels = int(360//args.grid_size)
    # print(out_channels)
    if args.pretrained:
        print("=> using pre-trained model '{}'".format(args.arch))
        model = timm.models.create_model(args.arch, pretrained=True, num_classes=10)  # 15
        print("=> creating model '{}'".format(args.arch))
        model = timm.models.create_model(args.arch, pretrained=False, num_classes=10)

    if not torch.cuda.is_available():
        print('using CPU, this will be slow')
    elif args.distributed:
        # For multiprocessing distributed, DistributedDataParallel constructor
        # should always set the single device scope, otherwise,
        # DistributedDataParallel will use all available devices.
        if args.gpu is not None:
            # When using a single GPU per process and per
            # DistributedDataParallel, we need to divide the batch size
            # ourselves based on the total number of GPUs of the current node.
            args.batch_size = int(args.batch_size / ngpus_per_node)
            args.workers = int((args.workers + ngpus_per_node - 1) / ngpus_per_node)
            model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu])
            # DistributedDataParallel will divide and allocate batch_size to all
            # available GPUs if device_ids are not set
            model = torch.nn.parallel.DistributedDataParallel(model)
    elif args.gpu is not None:
        model = model.cuda(args.gpu)
        # DataParallel will divide and allocate batch_size to all available GPUs
        # if args.arch.startswith('alexnet') or args.arch.startswith('vgg'):
        #     model.features = torch.nn.DataParallel(model.features)
        #     model.cuda()
        # else:
        # ckp = torch.load('checkpoint.pth.tar')
        # model.load_state_dict(ckp['state_dict'])
        model = torch.nn.DataParallel(model).cuda()

    # define loss function (criterion), optimizer, and learning rate scheduler
    criterion = nn.CrossEntropyLoss().cuda(args.gpu)

    optimizer = torch.optim.AdamW(model.parameters(),, weight_decay=args.weight_decay)
    """Sets the learning rate to the initial LR decayed by 10 every 30 epochs"""
    # scheduler = StepLR(optimizer, step_size=50, gamma=0.1)
    scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=args.epochs, eta_min=0)

    # optionally resume from a checkpoint
    if args.resume:
        if os.path.isfile(args.resume):
            print("=> loading checkpoint '{}'".format(args.resume))
            if args.gpu is None:
                checkpoint = torch.load(args.resume)
                # Map model to be loaded to specified single gpu.
                loc = 'cuda:{}'.format(args.gpu)
                checkpoint = torch.load(args.resume, map_location=loc)
            args.start_epoch = checkpoint['epoch']
            best_acc1 = checkpoint['best_acc1']
            if args.gpu is not None:
                # best_acc1 may be from a checkpoint from a different GPU
                best_acc1 =
            print("=> loaded checkpoint '{}' (epoch {})"
                  .format(args.resume, checkpoint['epoch']))
            print("=> no checkpoint found at '{}'".format(args.resume))

    cudnn.benchmark = True

    # Data loading code
    # traindir = os.path.join(, 'train')
    # valdir = os.path.join(, 'val')
    traindir =
    valdir = args.val_data
    normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                     std=[0.229, 0.224, 0.225])

    train_dataset = PoseData(

    if args.distributed:
        train_sampler =
        train_sampler = None

    train_loader =
        train_dataset, batch_size=args.batch_size, shuffle=(train_sampler is None),
        num_workers=args.workers, pin_memory=True, sampler=train_sampler)

    val_nuisances = ['shape', 'pose', 'texture', 'context', 'weather', 'occlusion']

    val_loaders = []
    for nuisance in val_nuisances:
                                PoseData(os.path.join(valdir, nuisance), transforms.Compose([
                                    transforms.Resize((224, 224)),
                                batch_size=args.batch_size, shuffle=False,
                                num_workers=args.workers, pin_memory=True,

    if args.evaluate:
        ckp = torch.load('model_best.pth.tar')
        # model.load_state_dict(torch.load('model_best.pth.tar'))
        validate(val_loaders, model, criterion, args)

    for epoch in range(args.start_epoch, args.epochs):
        if args.distributed:

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

        # evaluate on validation set
        acc1 = validate(val_loaders, model, criterion, args)


        # 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):
                'epoch': epoch + 1,
                'arch': args.arch,
                'state_dict': model.state_dict(),
                'best_acc1': best_acc1,
                'optimizer': optimizer.state_dict(),
                'scheduler': scheduler.state_dict()
            }, is_best)

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')
    top5 = AverageMeter('Acc@5', ':6.2f')
    progress = ProgressMeter(
        [batch_time, data_time, losses, top1, top5],
        prefix="Epoch: [{}]".format(epoch))

    # switch to train mode

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

        if torch.cuda.is_available():
            images = images.cuda(args.gpu, non_blocking=True)
            #images = rand_square_mask(images, mask_noise=True).cuda()
            target = target.cuda(args.gpu, non_blocking=True)

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

        # measure accuracy and record loss
        acc1, acc5 = accuracy(output, target, topk=(1, 5))
        losses.update(loss.item(), images.size(0))
        top1.update(acc1[0], images.size(0))
        top5.update(acc5[0], images.size(0))

        # compute gradient and do SGD step

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

        if i % args.print_freq == 0:

def validate(val_loaders, model, criterion, args):
    overall_top1 = 0
    for nuisance, val_loader in val_loaders:
        batch_time = AverageMeter('Time', ':6.3f', Summary.NONE)
        losses = AverageMeter('Loss', ':.4e', Summary.NONE)
        top1 = AverageMeter('Acc@1', ':6.2f', Summary.AVERAGE)
        top5 = AverageMeter('Acc@5', ':6.2f', Summary.AVERAGE)
        progress = ProgressMeter(
            [batch_time, losses, top1, top5],
            prefix=f'Test {nuisance}: ')

        # switch to evaluate mode

        with torch.no_grad():
            end = time.time()
            for i, data in enumerate(val_loader):
                # measure data loading time
                images = data['image']
                target = data['label']

                if torch.cuda.is_available():
                    images = images.cuda(args.gpu, non_blocking=True)
                    target = target.cuda(args.gpu, non_blocking=True)

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

                # measure accuracy and record loss
                acc1, acc5 = accuracy(output, target, topk=(1, 5))
                losses.update(loss.item(), images.size(0))
                top1.update(acc1[0], images.size(0))
                top5.update(acc5[0], images.size(0))

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

                if i % args.print_freq == 0:

        overall_top1 += top1.avg
    overall_top1 /= len(val_loaders)
    return top1.avg

def save_checkpoint(state, is_best, filename='{}_checkpoint.pth.tar'.format(args.arch)):, filename)
    if is_best:
        shutil.copyfile(filename, '{}_base_best.pth.tar'.format(args.arch))

class Summary(Enum):
    NONE = 0
    AVERAGE = 1
    SUM = 2
    COUNT = 3

class AverageMeter(object):
    """Computes and stores the average and current value"""

    def __init__(self, name, fmt=':f', summary_type=Summary.AVERAGE): = name
        self.fmt = fmt
        self.summary_type = summary_type

    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__)

    def summary(self):
        fmtstr = ''
        if self.summary_type is Summary.NONE:
            fmtstr = ''
        elif self.summary_type is Summary.AVERAGE:
            fmtstr = '{name} {avg:.3f}'
        elif self.summary_type is Summary.SUM:
            fmtstr = '{name} {sum:.3f}'
        elif self.summary_type is Summary.COUNT:
            fmtstr = '{name} {count:.3f}'
            raise ValueError('invalid summary type %r' % self.summary_type)

        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 display(self, batch):
        entries = [self.prefix + self.batch_fmtstr.format(batch)]
        entries += [str(meter) for meter in self.meters]

    def display_summary(self):
        entries = [" *"]
        entries += [meter.summary() for meter in self.meters]
        print(' '.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 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].reshape(-1).float().sum(0, keepdim=True)
            res.append(correct_k.mul_(100.0 / batch_size))
        return res

if __name__ == '__main__':

Test Code

from import Dataset, DataLoader
import pandas as pd
from torchvision import transforms
from PIL import Image
import torch
import torch.nn as nn
from glob import glob
from pathlib import PurePath
import numpy as np
import timm
import torchvision
import time

img_list = glob('/home/lynk/Documents/ROBINv1.1-cls-pose/iid_test/Images/*.jpg')

name_list = [

# conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 -c pytorch

class PoseData(Dataset):
    def __init__(self, transforms) -> None:
        the data folder should look like
        - datafolder
            - Images
            - labels.csv        
        self.img_list = glob('/home/lynk/Documents/ROBINv1.1-cls-pose/iid_test/Images/*.jpg')
        self.img_list = sorted(self.img_list, key=lambda x: PurePath(x).parts[-1][:-4])
        self.trs = transforms

    def __len__(self):
        return len(self.img_list)

    def __getitem__(self, index):
        image_dir = self.img_list[index]
        image_name = PurePath(image_dir).parts[-1]
        image =
        image = self.trs(image)

        return image, image_name

if __name__ == "__main__":
    normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                     std=[0.229, 0.224, 0.225])
    tfs = transforms.Compose([
        transforms.Resize((224, 224)),

    model = timm.models.resnet152(pretrained=True, num_classes=10)
    model = torch.nn.DataParallel(model)
    model.load_state_dict(torch.load('resnet152_base_best.pth.tar')['state_dict'], strict=False)
    model = model.cuda()

    dataset = PoseData(tfs)
    loader = DataLoader(dataset, batch_size=128, shuffle=False, drop_last=False, num_workers=4)

    image_dir = []
    preds = []
    for image, pth in loader:
        image = image.cuda()

        with torch.no_grad():

            pred = model(image)

            pred = torch.softmax(pred[:, :10], dim=1)
            pred = torch.argmax(pred[:, :10], dim=1)

            p = []
            for i in range(pred.size(0)):
        p = np.array(p)

    image_dir = np.array(sum(image_dir, []))
    preds = np.concatenate(preds)

    csv = {'imgs': np.array(image_dir), 'pred': np.array(preds),
    csv = pd.DataFrame(csv)

    csv.to_csv('evaluation/cls_ref/res/iid.csv', index=False)

Performance Result

Current iid performance:  0.08704620462046204
TOP-1@shape: 0.09378185524974515
TOP-1@pose: 0.04837209302325581
TOP-1@texture: 0.21238938053097345
TOP-1@context: 0.12522045855379188
TOP-1@weather: 0.12314540059347182
TOP-1@occlusion: 0.058660667935629004
Mean-TOP-1:  0.11026164264781119