Hello, I use a simple CNN with 70000 parameters and I use the Tiny-Imagenet dataset
my preprocessing is like below
train_batch_size = 128
test_batch_size = 128
lr = 0.001
momentum = 0.9
weight_decay = 0
seed = 7
margin = 1.0
log_interval = 10
resume = '-r'
epochs = 10
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])
transform_train = transforms.Compose([
transforms.Resize(32),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize])
transform_test = transforms.Compose([
transforms.Resize(32),
transforms.ToTensor(),
normalize])
train_dataset = datasets.ImageFolder('./tiny-imagenet-200/train', transform=transform_train)
test_dataset = datasets.ImageFolder('./tiny-imagenet-200/val', transform=transform_test)
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size = train_batch_size, shuffle = True, num_workers=0)
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size = test_batch_size, shuffle = False, num_workers=0)
after 100 epochs accuracy is equal to 23% and loss is 3% my optimizer is SGD. I used a deeper network like Alesxnet but my result didn’t change.
what’s wrong with my training. Any help would be appreciated.