Download presentation
Presentation is loading. Please wait.
Published byHarry Bradley Modified over 9 years ago
1
1 Andrew Ng, Associate Professor of Computer Science Robots and Brains
2
2 Who wants a robot to clean your house? [Photo Credit: iRobot]
3
3 Stanford STAIR Robot [Credit: Ken Salisbury]
4
4
5
5 What’s missing? Control Perception The software
6
6 Stanford autonomous helicopter
7
7 Computer GPS Accelerometers Compass
8
8
9
9 Computer program to fly helicopter [Courtesy of David Shim]
10
10 Option 1 BLACK
11
11 Machine learning Option 2
12
12 Machine learning
13
13 Machine learning to fly helicopter
14
14 What’s missing? The software Control Perception
15
15 “Robot, please find my coffee mug”
16
16 “Robot, please find my coffee mug” Mug
17
17 Why is computer vision hard? But the camera sees this:
18
18 Computer programs (features) for vision SIFT Spin image HoG Textons Shape context GIST
19
19 Why is speech recognition hard? What a microphone records: “Robot, please find my coffee mug.”
20
20 Computer programs (features) for audio ZCR Spectrogram MFCC Rolloff Flux
21
21 The idea: Most of perception in the brain may be one simple program
22
22 Auditory cortex learns to see Auditory Cortex The “one program” hypothesis [Roe et al., 1992]
23
23 Somatosensory cortex learns to see The “one program” hypothesis Somatosensory Cortex [Roe et al., 1992]
24
24 Neurons in the brain
25
25 Neural Network x1x1 x2x2 x3x3 Output Layer L 1 Layer L 2 Layer L 4 Layer L 3 x4x4
26
26 How does the brain process images? Neuron #1 of visual cortex (model) Neuron #2 of visual cortex (model) Primary visual cortex looks for “edges.”
27
27 Comparing to Biology Learning algorithm Visual cortex [PICTURE]
28
28 Comparing to Biology Learning algorithm Auditory cortex [PICTURE]
29
29 Computer vision results (NORB benchmark) Neural Network: accuracy Classical computer vision (SVM): accuracy
30
30 Missed Mugs True positivesFalse positives
31
31 Missed Mugs True positivesFalse positives
32
32 Missed Mugs True positivesFalse positives
33
33 Missed Mugs True positivesFalse positives
34
34 Missed Mugs True positivesFalse positives Results using non-embodied vision
35
35 Missed Mugs True positivesFalse positives
36
36 Missed Mugs True positivesFalse positives Results using non-embodied vision
37
37 Missed Mugs True positivesFalse positives Classifications using embodied agent
38
38 Missed Mugs True positivesFalse positives
39
39 Missed Mugs True positivesFalse positives Results using non-embodied vision
40
40 Missed Mugs True positivesFalse positives
41
41 Missed Mugs True positivesFalse positives
42
42 Hope of progress in Artificial Intelligence Email: ang@cs.stanford.edu
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.