Purposive visual imitation for learning structured tasks from videos
Abstract/Contents
- Abstract
- We are interested in robots that can learn structured tasks automatically from video demonstrations. These structured tasks, like cooking, are hierarchical and often long-horizon. This makes it hard to learn structured tasks purely based on interactions and first-person demonstrations. We advocate the advantages of learning from video demonstrations in this case, which is much easier to collect. The next question is: what exactly should we learn from videos? We propose Purposive Visual Imitation, where we learn the goal and intention of the demonstrator from videos. We show that this leads to stronger generalization than learning low-level trajectories and sequences of actions from the videos. In this dissertation, we will discuss how we can extract task graphs from online instructional videos, and how having such a graphical representation is beneficial for task execution. We will further introduce a motion reasoning framework to extract subgoals from video demonstrations, and how we can use planning to achieve these subgoals based on continuous inputs
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2020; ©2020 |
| Publication date | 2020; 2020 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Huang, De-An |
|---|---|
| Degree supervisor | Li, Fei Fei, 1976- |
| Thesis advisor | Li, Fei Fei, 1976- |
| Thesis advisor | Brunskill, Emma |
| Thesis advisor | Kochenderfer, Mykel J, 1980- |
| Degree committee member | Brunskill, Emma |
| Degree committee member | Kochenderfer, Mykel J, 1980- |
| Associated with | Stanford University, Computer Science Department. |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | De-An Huang |
|---|---|
| Note | Submitted to the Computer Science Department |
| Thesis | Thesis Ph.D. Stanford University 2020 |
| Location | electronic resource |
Access conditions
- Copyright
- © 2020 by De-An Huang
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...