| _version_ |
1826204297796255744
|
| collection |
MIT
|
| description |
© 2019 IEEE. Training models to high-end performance requires availability of large labeled datasets, which are expensive to get. The goal of our work is to automatically synthesize labeled datasets that are relevant for a downstream task. We propose Meta-Sim, which learns a generative model of synthetic scenes, and obtain images as well as its corresponding ground-truth via a graphics engine. We parametrize our dataset generator with a neural network, which learns to modify attributes of scene graphs obtained from probabilistic scene grammars, so as to minimize the distribution gap between its rendered outputs and target data. If the real dataset comes with a small labeled validation set, we additionally aim to optimize a meta-objective, i.e. downstream task performance. Experiments show that the proposed method can greatly improve content generation quality over a human-engineered probabilistic scene grammar, both qualitatively and quantitatively as measured by performance on a downstream task.
|
| first_indexed |
2024-09-23T12:52:04Z
|
| format |
Article
|
| id |
mit-1721.1/137178
|
| institution |
Massachusetts Institute of Technology
|
| language |
English
|
| last_indexed |
2024-09-23T12:52:04Z
|
| publishDate |
2021
|
| publisher |
IEEE
|
| record_format |
dspace
|
| spelling |
mit-1721.1/1371782021-11-04T03:21:25Z Meta-Sim: Learning to Generate Synthetic Datasets © 2019 IEEE. Training models to high-end performance requires availability of large labeled datasets, which are expensive to get. The goal of our work is to automatically synthesize labeled datasets that are relevant for a downstream task. We propose Meta-Sim, which learns a generative model of synthetic scenes, and obtain images as well as its corresponding ground-truth via a graphics engine. We parametrize our dataset generator with a neural network, which learns to modify attributes of scene graphs obtained from probabilistic scene grammars, so as to minimize the distribution gap between its rendered outputs and target data. If the real dataset comes with a small labeled validation set, we additionally aim to optimize a meta-objective, i.e. downstream task performance. Experiments show that the proposed method can greatly improve content generation quality over a human-engineered probabilistic scene grammar, both qualitatively and quantitatively as measured by performance on a downstream task. 2021-11-03T14:07:51Z 2021-11-03T14:07:51Z 2020-10 2021-04-15T17:07:08Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/137178 2020. "Meta-Sim: Learning to Generate Synthetic Datasets." Proceedings of the IEEE International Conference on Computer Vision, 2019-October. en 10.1109/ICCV.2019.00465 Proceedings of the IEEE International Conference on Computer Vision Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf IEEE arXiv
|
| spellingShingle |
Meta-Sim: Learning to Generate Synthetic Datasets
|
| title |
Meta-Sim: Learning to Generate Synthetic Datasets
|
| title_full |
Meta-Sim: Learning to Generate Synthetic Datasets
|
| title_fullStr |
Meta-Sim: Learning to Generate Synthetic Datasets
|
| title_full_unstemmed |
Meta-Sim: Learning to Generate Synthetic Datasets
|
| title_short |
Meta-Sim: Learning to Generate Synthetic Datasets
|
| title_sort |
meta sim learning to generate synthetic datasets
|
| url |
https://hdl.handle.net/1721.1/137178
|