Hunt for Dinosaurs with Intel® AI Technologies

This innovative workshop guides you through the AI techniques to create a likelihood map of discovering dinosaur bones in different locations. Explore machine learning algorithms powered by oneAPI software, including Intel® Extension for Scikit-learn*, NumPy, and Intel® Distribution of OpenVINO™ toolkit.

The workshop covers:

• Use scikit-learn* to preprocess metadata and experiment with semisupervised learning and clustering techniques, focusing on known dinosaur bone beds.
• Apply metadata labeling to aerial photos, which are fed to a convolutional neural network that uses the Intel® oneAPI Deep Neural Network Library (oneDNN) library to classify the aerial images.
• Use the Intel Distribution OpenVINO toolkit to access the inference capabilities of Intel CPUs. Use that inferencing capability to then process thousands of images and create a bone likelihood map. This map can be applied to other regions of the globe.

Note Because dinosaur bones on US Federal lands are protected, the actual map is blurred to maintain privacy.

0:00 Introduction

2:18 Discovery video

4:55 Set up in Intel® Developer Cloud

10:25 How do you know if it is a dinosaur bone?

13:10 Not all fossils are bone

14:07 View a depositional environment from the ground

17:25 View a depositional environment from the air

17:50 Where do you find dinosaur bones?

18:25 Finish a GitHub* clone

21:10 Create a dinosaur site treasure map

28:05 Live demo

28:33 Syllabus lab

32:52 Context matters with clustering

41:40 Collect aerial photos

43:00 Build, train, and score a model with PyTorch*

46:26 Fine tuning

48:52 Define utility functions

50:28 Define dataset transforms for training and validation sets

52:52 Create a dataset for training and validation

55:22 Dinosaur bone finder class

1:05:20 Load a saved model

1:10:32 Aerial labeling

1:11:40 Intel® AI software portfolio

1:12:36 OpenVINO toolkit

1:16:05 The meaning of "VINO"

1:16:45 Optimize an industry framework

1:18:45 The Model Optimizer tool

1:20:10 Post-Training Optimization Tool (POT)

1:20:50 Supported devices for easy deployment

1:21:05 An automatic device for easy deployment

1:22:32 Demo of finding dinosaur bones

1:25:00 Convert a model with ONNX* (Open Neural Network Exchange)

1:31:05 Generate a heatmap

1:33:30 Create and field test an AI map

1:36:20 Depositional environment differences

1:39:53 Obey all laws

1:42:15 Take a survey

1:46:26 Q&A