pytorch tnt meters使用
import math
import torch
import torchnet.meter as meter
import numpy as np m = meter.AverageValueMeter()
for i in range(1, 10):
m.add(i)
mean, std = m.value()mean,std(5.0, 2.7386127875258306)
m = meter.AverageValueMeter()
for i in range(1, 11):
m.add(i * i, n=i)
mean, std = m.value()mean,std(7.0, 19.939569196777708)
mtr = meter.ClassErrorMeter(topk=[1])
output = torch.eye(3)
if hasattr(torch, "arange"):
target = torch.arange(0, 3)
else:
target = torch.range(0, 2)
mtr.add(output, target)
err = mtr.value()target 0
1
2
[torch.FloatTensor of size 3]
output 1 0 0
0 1 0
0 0 1
[torch.FloatTensor of size 3x3]
err[0.0]
target[0] = 1
target[1] = 0
target[2] = 0
mtr.add(output, target)
err = mtr.value()target 1
0
0
[torch.FloatTensor of size 3]
output 1 0 0
0 1 0
0 0 1
[torch.FloatTensor of size 3x3]
err[50.0]
mtr = meter.ConfusionMeter(k=3)
output = torch.Tensor([[.8, 0.1, 0.1], [10, 11, 10], [0.2, 0.2, .3]])
if hasattr(torch, "arange"):
target = torch.arange(0, 3)
else:
target = torch.range(0, 2)
mtr.add(output, target)output 0.8000 0.1000 0.1000
10.0000 11.0000 10.0000
0.2000 0.2000 0.3000
[torch.FloatTensor of size 3x3]
output.shapetorch.Size([3, 3])
target 0
1
2
[torch.FloatTensor of size 3]
mtr.value()array([[1, 0, 0],
[0, 1, 0],
[0, 0, 1]], dtype=int32)
mtr = meter.ConfusionMeter(k=4, normalized=True)
output = torch.Tensor([
[.8, 0.1, 0.1, 0],
[10, 11, 10, 0],
[0.2, 0.2, .3, 0],
[0, 0, 0, 1],
])
target = torch.Tensor([0, 1, 2, 3])
mtr.add(output, target)
conf_mtrx = mtr.value()conf_mtrxarray([[1., 0., 0., 0.],
[0., 1., 0., 0.],
[0., 0., 1., 0.],
[0., 0., 0., 1.]], dtype=float32)
output 0.8000 0.1000 0.1000 0.0000
10.0000 11.0000 10.0000 0.0000
0.2000 0.2000 0.3000 0.0000
0.0000 0.0000 0.0000 1.0000
[torch.FloatTensor of size 4x4]
target[0] = 1
target[1] = 0
target[2] = 0
target[3]=2
mtr.reset()
mtr.add(output, target)
conf_mtrx = mtr.value()target 1
0
0
2
[torch.FloatTensor of size 4]
conf_mtrxarray([[0. , 0.5, 0.5, 0. ],
[1. , 0. , 0. , 0. ],
[0. , 0. , 0. , 1. ],
[0. , 0. , 0. , 0. ]], dtype=float32)
output 0,1,2,3，可以看出，target为0的时候有俩个，其中有一个被判为1一个被判为2，所以都是0.5
conf_mtrx.ravel()array([0. , 0.5, 0.5, 0. , 1. , 0. , 0. , 0. , 0. , 0. , 0. , 1. , 0. ,
0. , 0. , 0. ], dtype=float32)
ravel是拆开的意思，2*2的混淆矩阵的时候，拆开分别为tn,fp,fn,tp,这时1代表正样本
Confustion matrix of K rows and K columns, where rows corresponds
to ground-truth targets and columns corresponds to predicted
targets.
mtr = meter.AUCMeter()
test_size = 1000
mtr.add(torch.rand(test_size), torch.zeros(test_size))
mtr.add(torch.rand(test_size), torch.Tensor(test_size).fill_(1))
val, tpr, fpr = mtr.value()val,tpr,fpr(0.505625,
array([0. , 0.001, 0.001, ..., 0.998, 0.999, 1. ]),
array([0. , 0. , 0.001, ..., 1. , 1. , 1. ]))
mtr.reset()
mtr.add(torch.Tensor(test_size).fill_(0), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.1), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.2), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.3), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.4), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(1),
torch.Tensor(test_size).fill_(1))
val, tpr, fpr = mtr.value()val,tpr,fpr(1.0,
array([0. , 0.001, 0.002, ..., 1. , 1. , 1. ]),
array([0. , 0. , 0. , ..., 0.9996, 0.9998, 1. ]))
mtr.reset()
mtr.add(torch.Tensor(test_size).fill_(0), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.1), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.2), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.3), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.4), torch.Tensor(test_size).fill_(1))
mtr.add(torch.Tensor(test_size).fill_(1),
torch.Tensor(test_size).fill_(1))
val, tpr, fpr = mtr.value()val,tpr,fpr(1.0,
array([0.e+00, 5.e-04, 1.e-03, ..., 1.e+00, 1.e+00, 1.e+00]),
array([0. , 0. , 0. , ..., 0.9995 , 0.99975, 1. ]))
mtr.reset()
mtr.add(torch.Tensor(test_size).fill_(0), torch.Tensor(test_size).fill_(1))
mtr.add(torch.Tensor(test_size).fill_(0.1), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.2), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.3), torch.zeros(test_size))
mtr.add(torch.Tensor(test_size).fill_(0.4), torch.Tensor(test_size).fill_(1))
mtr.add(torch.Tensor(test_size).fill_(1),
torch.Tensor(test_size).fill_(1))
val, tpr, fpr = mtr.value()val,tpr,fpr(0.6666666666666667,
array([0.00000000e+00, 3.33333333e-04, 6.66666667e-04, ...,
9.99333333e-01, 9.99666667e-01, 1.00000000e+00]),
array([0., 0., 0., ..., 1., 1., 1.]))
mtr = meter.APMeter()
target = torch.Tensor([0, 1, 0, 1])
output = torch.Tensor([0.1, 0.2, 0.3, 4])
weight = torch.Tensor([0.5, 1.0, 2.0, 0.1])
mtr.add(output, target, weight)
ap = mtr.value()ap 0.6774
[torch.FloatTensor of size 1]
mtr = meter.APMeter()
target = torch.Tensor([0, 1, 0, 1])
output = torch.Tensor([0.1, 5, 0.3, 4])
weight = torch.Tensor([0.5, 1.0, 2.0, 0.1])
mtr.add(output, target, weight)
ap = mtr.value()ap 1
[torch.FloatTensor of size 1]
mtr = meter.mAPMeter()
target = torch.Tensor([0, 1, 0, 1])
output = torch.Tensor([0.1, 0.2, 0.3, 4])
weight = torch.Tensor([0.5, 1.0, 2.0, 0.1])
mtr.add(output, target)
ap = mtr.value()ap0.8333333730697632
val = (1 * 1.0 / 1.0 + 0 * 1.0 / 2.0 +
2.0 * 1.0 / 3.0 + 0 * 1.0 / 4.0) / 2.0平均精度就是各个recall下的平均精度，但是算着感觉不大对啊？
http://nooverfit.com/wp/david9%E7%9A%84%E6%99%AE%E5%8F%8A%E8%B4%B4%EF%BC%9A%E6%9C%BA%E5%99%A8%E8%A7%86%E8%A7%89%E4%B8%AD%E7%9A%84%E5%B9%B3%E5%9D%87%E7%B2%BE%E5%BA%A6ap-%E5%B9%B3%E5%9D%87%E7%B2%BE%E5%BA%A6%E5%9D%87/
val0.8333333333333333