SYNTHETIC DATA / AI Rev Lebaredian - Vice President, Simulation - - PowerPoint PPT Presentation
SYNTHETIC DATA / AI Rev Lebaredian - Vice President, Simulation Technolgy NEURAL NETWORKS NEED DATA And Labels! Deep Learning is amazing! Require huge amounts of quality data Data needs labeling For some problems, data + labeling is available
Rev Lebaredian - Vice President, Simulation Technolgy
2
Deep Learning is amazing! Require huge amounts of quality data Data needs labeling For some problems, data + labeling is available Good data doesn’t exists for most problems
Image-Net (http://www.image-net.org/) example
3
4
5
7
Real data is expensive, sometimes dangerous Synthetic labels are automatic and accurate Useful for validation, in addition to training
8
Adrien Gaidon, Qiao Wang, Yohann Cabon, Eleonora Vig: Virtual Worlds as Proxy for Multi-Object Tracking Analysis IEEE CVPR 2016
9
10
11
12
SIMULATE TRAIN MODELS COLLECT DATA DRIVE
Lanes Lights
Path
Signs
Pedestrians Cars
Lanes Lights
Path
Signs
Pedestrians Cars
13
14
15
AlphaZero OpenAI Five OpenAI, 2018 Deepmind, 2018
16
Locomotion/Animation
Liang, Makoviychuk, Handa etc, 2018 NVIDIA
17
Sim2Real Robotics
Chebotar, Handa, Makoviychuk, etc, 2018 NVIDIA
18
19
20
21
22
23
propagation between different phases
24
25
LEARN & PLAN SEE, HEAR, TOUCH ACT
26
27
28
29
ISAAC SIM
30
31
32
33