Meta-Sim: Learning to Generate Synthetic Datasets
This addresses the high cost of labeled data for training models, offering a domain-specific solution for synthetic data generation.
The authors tackled the problem of expensive labeled datasets by proposing Meta-Sim, a method to automatically generate synthetic datasets that minimize distribution gaps with real data and optimize downstream task performance, showing great improvements in content generation quality over human-engineered methods.
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.