linquanxu / feature_extract-from-caffe

Use python script to extract feature from trained caffe-net,images and file-label pair text in the form of 1-D array.

feature_extract-from-caffe

Use python script to extract feature from trained caffe-net,images and file-label pair text in the form of 1-D array.

Sometimes after you train a caffe net,you probably want to extract features from different layers and feed them to classifiers such as SVM or RF, but the built-in tools of caffe for feature_extraction is hard to use.Thus this python script is developed to help you extract feature from caffemodel,original images,image-label-pair-text,and save the feature and label in the form of npy file(which can be load by numpy function numpy.load() as a form of nd-array),and then you can use sklearn.svm to train a classifier use the feature and labels you extracted.

---

FILE NEEDED: a already trained caffe-model the deploy file of the caffe-model the original image of your dataset the image-label pair text file

---

Dependency: caffe,numpy