Simple implementation of (convolutional) autoencoders in Keras and Tensorflow.
The code is commented and pretty much self-explanatory.
Data stores
I
: 1000 flattened images,IG
: same images converted to grayscale,h
andw
: height and width (484x644),I2
,IG2
,w2
andh2
: same as above but images are scaled and cropped to 64x64.