Automatic Generation of Dense Non-rigid Optical Flow
This addresses the problem of dataset scarcity for non-rigid optical flow in computer vision, enabling better training for flow estimation models, though it is incremental as it builds on existing object extraction and deformation techniques.
The paper tackles the lack of large-scale datasets for dense non-rigid optical flow in real-world imagery by proposing an automatic generation framework that extracts and matches objects from videos to compute dense flow fields, and shows that training on this data outperforms methods trained on rigid synthetic data with models like FlowNet-S, LiteFlowNet, PWC-Net, and RAFT.
There hardly exists any large-scale datasets with dense optical flow of non-rigid motion from real-world imagery as of today. The reason lies mainly in the required setup to derive ground truth optical flows: a series of images with known camera poses along its trajectory, and an accurate 3D model from a textured scene. Human annotation is not only too tedious for large databases, it can simply hardly contribute to accurate optical flow. To circumvent the need for manual annotation, we propose a framework to automatically generate optical flow from real-world videos. The method extracts and matches objects from video frames to compute initial constraints, and applies a deformation over the objects of interest to obtain dense optical flow fields. We propose several ways to augment the optical flow variations. Extensive experimental results show that training on our automatically generated optical flow outperforms methods that are trained on rigid synthetic data using FlowNet-S, LiteFlowNet, PWC-Net, and RAFT. Datasets and implementation of our optical flow generation framework are released at https://github.com/lhoangan/arap_flow