caffe_confusion.py

import numpy as np
from sklearn.metrics import confusion_matrix
preds = ()
labels = ()
n_samples = ... number of examples in test set ...
batch_size = ... batch size ...
net = ... caffe test net ...
n_batch = int(np.ceil(n_samples / batch_size))
for i in range(n_batch):
  net.forward()
  score, label = net.blobs['fc8'], net.blobs['label']
  pred = score.argmax(axis=1)
  preds += pred,
  labels ++ label,
preds = np.vstack(preds)[:n_samples]
labels = np.vstack(labels)[:n_samples]
cm = confusion_matrix(labels.flatten(), preds.flatten())
plt.figure()
plt.imshow(cm, interpolation='nearest', cmap=plt.jet)
plt.show()