It also looks good to save conv_inv.

[dhkdnduq]

main.py

origin code :
for i in range(H * W):
# cov[:, :, i] = LedoitWolf().fit(embedding_vectors[:, :, i].numpy()).covariance_
cov[:, :, i] = np.cov(embedding_vectors[:, :, i].numpy(), rowvar=False) + 0.01 * I
# save learned distribution
train_outputs = [mean, cov]

change to :
for i in range(H * W):
# cov[:, :, i] = LedoitWolf().fit(embedding_vectors[:, :, i].numpy()).covariance_
cov[:, :, i] = np.cov(embedding_vectors[:, :, i].numpy(), rowvar=False) + 0.01 * I
# save learned distribution
conv_inv = np.linalg.inv(cov.T).T
train_outputs = [mean, cov, conv_inv]

how to using :

dist_list = []
for i in range(H * W):
mean = train_outputs[0][:, i]
#conv_inv = np.linalg.inv(train_outputs[1][:, :, i])
dist = [mahalanobis(sample[:, i], mean, train_outputs[2][:, :, i]) for sample in embedding_vectors]
dist_list.append(dist)

in my opinion

[DeepKnowledge1]

you do not need to save cov as well

//save only
train_outputs = [mean, conv_inv]