What Makes Good Synthetic Training Data for Learning Disparity and Optical Flow Estimation?

TL;DR

This paper explores how to generate effective synthetic training data for disparity and optical flow estimation, emphasizing dataset properties and training schedules to improve network performance and generalization.

Contribution

It introduces methods for creating synthetic data tailored for disparity and optical flow tasks and evaluates their impact on learning outcomes.

Findings

Synthetic data quality significantly affects model performance.

Learning schedules with varied data types enhance generalization.

Dataset properties influence training effectiveness.

Abstract

The finding that very large networks can be trained efficiently and reliably has led to a paradigm shift in computer vision from engineered solutions to learning formulations. As a result, the research challenge shifts from devising algorithms to creating suitable and abundant training data for supervised learning. How to efficiently create such training data? The dominant data acquisition method in visual recognition is based on web data and manual annotation. Yet, for many computer vision problems, such as stereo or optical flow estimation, this approach is not feasible because humans cannot manually enter a pixel-accurate flow field. In this paper, we promote the use of synthetically generated data for the purpose of training deep networks on such tasks.We suggest multiple ways to generate such data and evaluate the influence of dataset properties on the performance and generalization properties of the resulting networks. We also demonstrate the benefit of learning schedules that use different types of data at selected stages of the training process.