wandb pytorch:每个类的top1精度

Question

我在验证集中有5个类，我想使用wandb在验证循环中根据每个类的top1结果绘制一个图。我已经尝试了基于5个类的平均值的单个精度图，它工作得很好，但我想为每个类做一个单独的方式，比如top1精度。我做不到，有什么办法可以做到吗?

验证装载机

val_loaders = []
for nuisance in val_nuisances:
val_loaders.append((nuisance, torch.utils.data.DataLoader(
datasets.ImageFolder(os.path.join(valdir, nuisance), transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
normalize,
])),
batch_size=args.batch_size, shuffle=False,
num_workers=args.workers, pin_memory=True,
)))

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

验证循环

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(
len(val_loader),
[batch_time, losses, top1, top5],
prefix=f'Test {nuisance}: ')
# 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)
if torch.cuda.is_available():
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:
progress.display(i)
progress.display_summary()
overall_top1 += top1.avg
overall_top1 /= len(val_loaders)
return top1.avg

Answer 1

我在代码中没有看到任何到W&B的日志，但是记录每个类的top1精度将只是

class_names = ['shape', 'pose', 'texture', 'context', 'weather']
top1_accuracies = [0.9, 0.8, 0.9, 0.9, 0.8]
wandb.log({class_names[0]: top1_accuracies[0], class_names[1]: top1_accuracies[1], ...}

在上面的示例中，看起来您实际上并没有为每个类的top1精度创建变量。你会想先这么做。摘自https://stackoverflow.com/a/50977153/3959708

你可以使用sklearn的混淆矩阵来获得准确率

from sklearn.metrics import confusion_matrix
import numpy as np
y_true = [0, 1, 2, 2, 2]
y_pred = [0, 0, 2, 2, 1]
target_names = ['class 0', 'class 1', 'class 2']
#Get the confusion matrix
cm = confusion_matrix(y_true, y_pred)
#array([[1, 0, 0],
#   [1, 0, 0],
#   [0, 1, 2]])
#Now the normalize the diagonal entries
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
#array([[1.        , 0.        , 0.        ],
#      [1.        , 0.        , 0.        ],
#      [0.        , 0.33333333, 0.66666667]])
#The diagonal entries are the accuracies of each class
cm.diagonal()
#array([1.        , 0.        , 0.66666667])