at runtime image classification is taking long time

[vijay krishna]

Hi all,

Am able to generate eleb-classifier.nn4.small2.v1.pkl based on the example(https://cmusatyalab.github.io/openface/demo-3-classifier/)

given. The .pkl seems to be around 4 GB (for 13k photos)
In future i need to train with 100,000 photos.Once i 
deploy the application(.pkl file) in production, classifying
each image with pkl file at runtime would take longer time.
What is the best way to implement the solution in production so that each 
image classification should complete within earliest time possible. 

Thanks
vijay