Making plots using visdom
It is used to realise the Confusion Matrix. The diagonal elements in the matrix represent the number of correctly classified classes but in the case of visdom plotter you will have lower diagonal for it.
Without Flipping(lower diagonal will correspond to correctly classified classes)
import torchnet as tnt
confusion_matrix = tnt.meter.ConfusionMeter(3) # 3 classes
y_actu = np.array([2, 2, 1, 2, 0, 2, 1])
y_pred = np.array([2, 2, 1, 2, 0, 2, 1])
confusion_matrix.add((torch.from_numpy(y_actu)),(torch.from_numpy(y_pred)))
print(confusion_matrix.value)
vis = visdom.Visdom(port='6002')
vis.heatmap(
X=confusion_matrix.value(),
opts=dict(
columnnames=['a', 'b', 'c'],
rownames=['a', 'b', 'c'],
colormap='Electric',
)
)
With Flipping(upper diagonal will correspond to correctly classified classes)
Replace the above with the below code.
vis.heatmap(
X=np.flipud(confusion_matrix.value()), # Flip the matrix
opts=dict(
columnnames=['a', 'b', 'c'],
rownames=['c', 'b', 'a'], # change the order of labels
colormap='Electric',
)
)
X = [[] for i in range(7)]
Y = [[] for i in range(7)]
trace=[]
colors = ["red","blue","yellow","green","orange","violet","brown"]
symbols = [5, 1, 31, 104, 126, 21, 27]
m=[1,2,3,4,5,6,7]
n =[8,9,10,11,12,13,14]
c = -1
for j in range(7):
print()
c=c+1
X[j].append(m[c])
Y[j].append(n[c])
tr = dict(x=X[j], y=Y[j], marker={'color': colors[j], 'symbol': symbols[j], 'size': "10"},
mode="markers+lines",
name='class-'+str(j), type="custom")
trace.append(tr)
layout=dict(title="XY---Curve", xaxis={'title':'X'}, yaxis={'title':'Y'})
vis._send({'data': trace, 'layout': layout, 'win': 'XYCURVE','update':'append'})
trace1 = dict(x=[1,3,5,7], y=[1,3,5,7], marker={'color': 'red', 'symbol': 5, 'size': "10"},
mode="markers+lines",
name='class-1', type="custom")
trace2 = dict(x=[2,4,6,8], y=[2,3,4,5], marker={'color': 'blue', 'symbol': 1, 'size': "10"},
mode="markers+lines",
name='class-2', type="custom")
trace3 = dict(x=[1,2,3,4], y=[4,3,2,1], marker={'color': 'green', 'symbol': 31, 'size': "10"},
mode="markers+lines",
name='class-3', type="custom")
trace4 = dict(x=[5,3,2,1], y=[1,2,3,6], marker={'color': 'orange', 'symbol': 104, 'size': "10"},
mode="markers+lines",
name='class-4', type="custom")
trace5 = dict(x=[3,6,9,10], y=[3,2,1,0], marker={'color': 'violet', 'symbol': 126, 'size': "10"},
mode="markers+lines",
name='class-5', type="custom")
trace6 = dict(x=[1,2,3,4], y=[1,4,7,8], marker={'color': 'brown', 'symbol': 21, 'size': "10"},
mode="markers+lines",
name='class-6', type="custom")
trace7 = dict(x=[1,2,6,8], y=[1,4,3,7], marker={'color': 'yellow', 'symbol': 27, 'size': "10"},
mode="markers+lines",
name='class-7', type="custom")
layout=dict(title="XY---Curve", xaxis={'title':'X'}, yaxis={'title':'Y'})
vis._send({'data': [trace1, trace2, trace3, trace4, trace5, trace6, trace7], 'layout': layout, 'win': 'XYCURVE','update':'append'})
vis_line = vis.line(X = np.array([1,2,3]), Y = np.array([[1,1,1],[2,2,4],[3,4,7]]), win="my_line")
def plot_combine(name,d):
#index = {}
#multiple plots in one single graph
X = []
Y = []
legend = []
for k, v in sorted(d.items()):
print(k,v[0],v[1])
Y.append(v[0])
X.append(v[1])
legend.append(k)
Y = np.array([Y])
X = np.array([X])
vis.line(
Y=Y,
X=X,
win=name,
env= 'main',
opts=dict(
title=name,
legend=legend
),
update= 'append',
)
for i in [1,2,3,4,5,6]:
d = {'class-1': [i+1,i*3] ,'class-2': [i*4,i+2]}
plot_combine("combined_Lines",d) 
def combined_plot():
x = [0,1,2,3,4]
data = [{
'x': x,
'y': list(map(lambda i: i, x)),
'marker':{
'color': 'red',
'symbol': 104,
'size': "10"},
'mode':"markers+lines",
'name':'class-1',
'type':'line',
},{
'x': x,
'y': list(map(lambda i: i*2, x)),
'marker':{
'color': 'green',
'symbol': 104,
'size': "10"},
'mode':"markers+lines",
'name':'class-2',
'type':'line',
},
{
'x': x,
'y': list(map(lambda i: i*3, x)),
'marker':{
'color': 'blue',
'symbol': 104,
'size': "10"},
'mode':"markers+lines",
'name':'class-3',
'type':'line',
}
,
{
'x': x,
'y': list(map(lambda i: i*4, x)),
'marker':{
'color': 'green',
'symbol': 104,
'size': "10"},
'mode':"markers+lines",
'name':'class-3',
'type':'line',
}
]
win = 'custom_multiple_lines'
env = 'main'
layout= {
'title':"combined_plot",
'xaxis':{'title':'x1'},
'yaxis':{'title':'x2'}
}
opts = {}
vis._send({'data': data, 'win': win, 'eid': env, 'layout': layout, 'opts': opts})
combined_plot()
img = np.random.rand(3, 512, 256)
vis.images(img, win="IMG", opts=dict(title="IMAGE"))
imgs = np.random.random((4, 3, 128, 128))*255
vis.images(
imgs,
opts=dict(title='Random Images', caption='How random.')
)
import matplotlib.pyplot as plt
plt.plot([1, 7, 2, 4])
plt.title("Line")
plt.ylabel('some numbers')
vis.matplot(plt)
# 2D scatterplot with custom intensities (red channel)
vis.scatter(
X=np.random.rand(255, 2),
Y=(np.random.rand(255) + 1.5).astype(int),
opts=dict(
markersize=10,
markercolor=np.random.randint(0, 255, (2, 3,)),
),
)
# pie chart
X = np.asarray([10, 22, 13, 52])
vis.pie(
X=X,
opts=dict(legend=['A','B','C','D'])
)