python如何计算两个二维特征向量相似度

Question

从两个图片提取的两个特征向量，二维结构，结构如下：
array([[ 2.62957041e+00, 7.46614219e-05, 2.37797423e-05, ...,
-4.94050192e-04, 2.06032040e-03, 4.94050192e-04],
[-1.05751487e+00, 0.00000000e+00, 0.00000000e+00, ...,
-4.91478900e-04, 1.09093972e-03, 5.50124164e-04],
[ 2.73112827e+00, -1.22879321e-03, -1.01920502e-03, ...,
-1.68389973e-06, 4.27874303e-06, 1.68389973e-06],
...,
[-2.34248195e+00, 0.00000000e+00, 0.00000000e+00, ...,
1.95515861e-03, 6.74179684e-03, 5.66436691e-03],
[-2.34357433e+00, -2.35404491e-03, 1.88434007e-03, ...,
0.00000000e+00, 0.00000000e+00, 0.00000000e+00],
[-2.56323038e+00, 0.00000000e+00, 0.00000000e+00, ...,
-2.82040412e-03, 1.13440457e-02, 5.12339584e-03]])
二维结构的使用余弦夹角计算不了，
dot(arr1,arr2)/(linalg.norm(arr1)*linalg.norm(arr2))
报错：
ValueError: shapes (1024,65) and (1024,65) not aligned: 65 (dim 1) != 1024 (dim 0)
有经验的帮忙看下！

Answer 1

ValueError: shapes (1024,65) and (1024,65) not aligned: 65 (dim 1) != 1024 (dim 0)
这个提示说明你的矩阵维度不匹配，不能相乘，如果第一个矩阵是1024x65，第二个必须是65x1024，你转置下看看

另外参考
https://blog.csdn.net/xuxiatian/article/details/91388480