1 of 27

TensorFlow with EE

Nick Clinton, Earth Engine Developer Relations

Chris Brown, Software Engineer

nclinton@google.com

https://goo.gl/kZmFfG

2 of 27

"Deep learning is a particular kind of machine learning that achieves great power and flexibility by representing the world as a nested hierarchy of concepts, with each concept defined in relation to simpler concepts, and more abstract representations computed in terms of less abstract ones."

Ian Goodfellow and Yoshua Bengio and Aaron Courville, "Deep Learning"

3 of 27

About TensorFlow

TensorFlow™ is an open source software library for high performance numerical computation. Its flexible architecture allows easy deployment of computation across a variety of platforms (CPUs, GPUs, TPUs), and from desktops to clusters of servers to mobile and edge devices. Originally developed by researchers and engineers from the Google Brain team within Google’s AI organization, it comes with strong support for machine learning and deep learning and the flexible numerical computation core is used across many other scientific domains.

https://www.tensorflow.org/

4 of 27

Disclaimers

  • This is not a deep learning course
  • This is not a TensorFlow course
  • These are new features! (And more to come!)

5 of 27

What this course IS about

  • An end-to-end workflow for using TensorFlow to perform classification on data from Earth Engine.
  • A crash course in the TensorFlow Estimator API for land cover classification.
  • A fancy teaser model for EE + TF

6 of 27

7 of 27

Useful docs

https://www.tensorflow.org/get_started/

​

https://github.com/tensorflow/tensorflow

​

https://cloud.google.com/ml-engine/docs/tensorflow/getting-started-training-prediction

8 of 27

https://colab.research.google.com/notebooks/welcome.ipynb

9 of 27

's

10 of 27

11 of 27

12 of 27

Search

13 of 27

14 of 27

15 of 27

  1. Get these slides https://goo.gl/kZmFfG
  2. Join group
  3. Make a copy of this notebook.
    1. (File > Save a copy in Drive)

16 of 27

Train/Test data (Export.table)

.TFRecord

Image data (Export.Image)

.TFRecord

.train()

.predict()

Predictions

.TFRecord

upload

17 of 27

18 of 27

19 of 27

20 of 27

So why would you use ?

21 of 27

Let's get spatial!

We can apply models with regions of interest larger/deeper than a pixel with an arbitrary level of complexity!

https://arxiv.org/abs/1411.4038

22 of 27

… and let's detect objects!

23 of 27

3 Dominating Flavors

  1. Models strided over a region

​

  • Fully Convolutional Neural Nets

​

  • Time series (RNN)

24 of 27

Convolution

25 of 27

http://vision03.csail.mit.edu/cnn_art/index.html#v_single

26 of 27

Full Convolutional Neural Nets

http://www.mdpi.com/2072-4292/10/1/52

27 of 27

Car Detector

Generate Training Data

Annotate examples

Export training data

Train Model

Export Inference Data

Predict

Extract Features