deeplearning-course-project

We are trying to learn a good metric of distance between images using deep convolutional neural netowrks, which can be then applied to image classification based on the classsic knn method.

数据部分

orls_faces文件夹是ORL face database 人脸识别数据集 the Olivetti Research Laboratory in Cambridge, UK 40个类别,每个类别10个样本 PGM格式,92x112的8位灰度图。

参考

F. Samaria and A. Harter "Parameterisation of a stochastic model for human face identification" 2nd IEEE Workshop on Applications of Computer Vision December 1994, Sarasota (Florida).

链接

http://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase.html

load_data.py

load_all_data() 返回所有的数据,返回数据是list,包含40个list元素,每个list包含同一个类别的10个np.matrix(112x92,dtype=uint8) [[np.matrix(112x92),...],...]