No content

Transfer Learning With TensorFlow — Re-using Powerful Models 

Why TensorFlow? Apart from the excellent functionalities, services and high-level ops for advanced computations, TensorFlow was built to be production ready — making scalability an easy-go! 

Production-ready A great example is how easy it is to re-use powerful models in TensorFlow. This technique in machine learning is called Transfer Learning. 

Transfer learning and domain adaptation refer to the situation where what has been learned in one setting (i.e., distribution P1) is exploited to improve generalization in another setting (say distribution P2). — Extracted from Deep Learning, pg 536 “ 

Transfer Learning is a technique where a model which has already been trained on one problem is used in some way on a second related problem. The advantages are numerous, inclusive of: ● Learning a new task relies on previously learned tasks ● Less training data is needed, hence learning process is faster ● Learning gets to be more accurate 

Transfer Learning for Image Classification Our codelab will cover Transfer Learning for Image Classification using several pre-trained powerful deep learning models built by researchers in the past. We’ll use the following pre-trained models: ● MobileNet ● ResNet ● Inception (aka GoogleNet) 

Meet TensorFlow Hub Quite interestingly, TensorFlow 2.x comes with a lot of these pre-trained models to be easily plugged in to your code and utilized. Amazing right? Meet tensorflow.org/hub 

Image Classification with normal CNN: bit.ly/tfdevsummit_codelab_cnn 

Image Classification with Transfer Learning (MobileNet, Inception, ResNet): bit.ly/tfdevsummit_codelab_tl 

@olayinkapeter_