AutoFlow: Learning a Better Training Set for Optical Flow
This addresses the challenge of manually creating synthetic datasets for optical flow, which is time-consuming and inflexible, by automating the process for improved model training.
The paper tackles the problem of generating synthetic training data for optical flow by introducing AutoFlow, a method that automatically renders data to optimize model performance on a target dataset, achieving state-of-the-art accuracy in pre-training PWC-Net and RAFT.
Synthetic datasets play a critical role in pre-training CNN models for optical flow, but they are painstaking to generate and hard to adapt to new applications. To automate the process, we present AutoFlow, a simple and effective method to render training data for optical flow that optimizes the performance of a model on a target dataset. AutoFlow takes a layered approach to render synthetic data, where the motion, shape, and appearance of each layer are controlled by learnable hyperparameters. Experimental results show that AutoFlow achieves state-of-the-art accuracy in pre-training both PWC-Net and RAFT. Our code and data are available at https://autoflow-google.github.io .