In [79]:

    

import numpy as np
import torch


    

In [131]:

    

yolo = 45
fastyolo = 155
faster_rcnn_vgg = 7
faster_rcnn_zf = 18
fast_rcnn = 0.5


    

In [138]:

    

(yolo / faster_rcnn_vgg + yolo / faster_rcnn_zf) / 2


    

    
Out[138]:





4.464285714285714

In [ ]:

    




    

In [90]:

    

inp = torch.Tensor(1,5,5,5).random_()


    

In [91]:

    

inp.shape


    

    
Out[91]:





torch.Size([1, 5, 5, 5])

In [92]:

    

inp


    

    
Out[92]:





(0 ,0 ,.,.) = 
  1.2248e+06  7.9921e+05  1.5734e+07  5.0970e+06  1.2984e+06
  9.1090e+06  1.6063e+07  1.5996e+07  9.2673e+06  1.0755e+07
  4.5775e+05  1.4311e+07  1.6090e+07  6.4259e+06  1.3480e+07
  2.4642e+06  1.0476e+07  8.6037e+06  8.9452e+06  1.0454e+07
  5.1703e+06  1.3040e+07  3.2060e+06  3.1052e+06  8.2502e+06

(0 ,1 ,.,.) = 
  3.3638e+06  4.6623e+06  1.0991e+07  1.3211e+07  5.0972e+06
  4.9520e+06  3.4779e+04  1.0938e+06  1.2874e+07  1.4750e+07
  4.2556e+06  2.1060e+06  1.3499e+07  2.5491e+06  1.2473e+07
  4.9722e+06  1.6460e+07  4.5339e+06  1.6667e+07  1.0846e+07
  7.9361e+06  1.2648e+05  6.7994e+06  1.4796e+07  1.4011e+07

(0 ,2 ,.,.) = 
  7.0099e+06  3.6675e+05  6.7827e+06  2.8746e+06  8.2469e+06
  1.3239e+07  2.4886e+05  1.3017e+07  8.3168e+06  3.5989e+06
  1.3399e+07  1.4565e+07  8.4372e+06  1.1126e+07  8.0760e+06
  7.4382e+06  9.6461e+06  1.6129e+07  1.6132e+07  7.1095e+06
  6.1901e+06  5.2182e+06  5.8249e+06  4.1629e+06  6.1138e+06

(0 ,3 ,.,.) = 
  1.4324e+07  1.5362e+07  1.6225e+07  8.7385e+06  3.4110e+05
  4.6880e+06  1.6047e+06  1.4028e+07  9.8412e+06  1.4859e+07
  1.0942e+07  8.4442e+06  1.6709e+07  6.8308e+06  1.2146e+07
  1.0931e+07  2.3206e+06  4.0802e+06  7.0302e+06  5.2488e+05
  6.5140e+06  4.4017e+06  1.1318e+07  5.4130e+06  1.4478e+07

(0 ,4 ,.,.) = 
  1.4875e+06  1.4074e+07  1.1100e+06  1.2165e+07  9.0916e+06
  8.0989e+06  8.2595e+06  5.5936e+06  6.6251e+06  3.6089e+06
  1.6621e+07  1.6777e+07  3.0415e+04  4.5341e+06  1.4909e+07
  5.4878e+05  5.9468e+05  7.9935e+06  7.2528e+06  1.3032e+07
  7.1992e+06  2.5797e+06  3.8977e+06  9.2879e+06  1.5335e+07
[torch.FloatTensor of size 1x5x5x5]

In [95]:

    

inp = torch.autograd.Variable(inp)


    

In [101]:

    

convlayer = torch.nn.Conv2d(5,4,5)


    

In [102]:

    

convlayer(inp)


    

    
Out[102]:





Variable containing:
(0 ,0 ,.,.) = 
1.00000e+06 *
   8.0586

(0 ,1 ,.,.) = 
1.00000e+06 *
   0.1184

(0 ,2 ,.,.) = 
1.00000e+06 *
   5.7983

(0 ,3 ,.,.) = 
1.00000e+06 *
   1.8350
[torch.FloatTensor of size 1x4x1x1]

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [25]:

    




    

In [76]:

    

# inp = torch.Tensor(10).random_()
# inp = torch.autograd.Variable(inp)

inp = torch.autograd.Variable(torch.Tensor(5,1,5,5).zero_())

convlayer = torch.nn.Conv2d(1,20,5)

out = convlayer(inp)

print(f'Input tensor dimensions:  {inp.size()}')
print(f'Performing Convolution:   {convlayer}')
print(f'Output tensor dimensions: {out.shape}')


    

    




Input tensor dimensions:  torch.Size([5, 1, 5, 5])
Performing Convolution:   Conv2d (1, 20, kernel_size=(5, 5), stride=(1, 1))
Output tensor dimensions: torch.Size([5, 20, 1, 1])

In [77]:

    

inp[:1]


    

    
Out[77]:





Variable containing:
(0 ,0 ,.,.) = 
  0  0  0  0  0
  0  0  0  0  0
  0  0  0  0  0
  0  0  0  0  0
  0  0  0  0  0
[torch.FloatTensor of size 1x1x5x5]

In [78]:

    

out[0]


    

    
Out[78]:





Variable containing:
(0 ,.,.) = 
 -0.1920

(1 ,.,.) = 
  0.0494

(2 ,.,.) = 
 -0.0975

(3 ,.,.) = 
 -0.1194

(4 ,.,.) = 
  0.0099

(5 ,.,.) = 
 -0.0099

(6 ,.,.) = 
 -0.0704

(7 ,.,.) = 
  0.1014

(8 ,.,.) = 
 -0.0377

(9 ,.,.) = 
  0.1925

(10,.,.) = 
 -0.1181

(11,.,.) = 
  0.1191

(12,.,.) = 
 -0.1801

(13,.,.) = 
  0.0055

(14,.,.) = 
  0.1220

(15,.,.) = 
 -0.0747

(16,.,.) = 
 -0.1595

(17,.,.) = 
 -0.0823

(18,.,.) = 
  0.0461

(19,.,.) = 
 -0.0982
[torch.FloatTensor of size 20x1x1]

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [ ]:

    




    

In [1]:

    

import numpy as np

from scipy.spatial.distance import euclidean


    

In [5]:

    

#              x1  y1  x2  y2
c1 = np.array([10, 10, 80, 80])
c2 = np.array([20, 20, 90, 90])


    

In [6]:

    

np.sqrt((c2[2]-c1[2])**2+(c2[0]-c1[0])**2 
       +(c2[3]-c1[3])**2+(c2[1]-c1[1])**2)


    

    
Out[6]:





20.0

In [7]:

    

euclidean(c1, c2)


    

    
Out[7]:





20.0

In [8]:

    

euclidean(c2, c1)


    

    
Out[8]:





20.0

In [ ]: