GMFlow: Learning Optical Flow via Global Matching

By Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, Dacheng Tao. Computer Vision and Pattern Recognition, 2021

544 citations recorded by the source corpus.

Learning-based optical flow estimation has been dominated with the pipeline of cost volume with convolutions for flow regression, which is inherently limited to local correlations and thus is hard to address the long-standing challenge of large displacements. To alleviate this, the state-of-the-art framework RAFT gradually improves its prediction quality by using a large number of iterative refinements, achieving remarkable performance but introducing linearly increasing inference time. To enable both high accuracy and efficiency, we completely revamp the dominant flow regression pipeline by reformulating optical flow as a global matching problem, which identifies the correspondences by directly comparing feature similarities. Specifically, we propose a GMFlow framework, which consists of three main components: a customized Transformer for feature enhancement, a correlation and softmax layer for global feature matching, and a self-attention layer for flow propagation. We further introduce a refinement step that reuses GMFlow at higher feature resolution for residual flow prediction. Our new framework outperforms 31-refinements RAFT on the challenging Sintel benchmark, while using only one refinement and running faster, suggesting a new paradigm for accurate and efficient optical flow estimation. Code is available at https://github.com/haofeixu/gmflow.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/e4733ff919aefc7048774876b05e73bae56fcb49

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Unifying Flow, Stereo and Depth Estimation (2022, methodology): “This not only reduces parameters but also improves generalization, as shown in the original GMFlow [38] paper.”
VideoComposer: Compositional Video Synthesis with Motion Controllability (2023, methodology): “In our implementation, we employ the widely used pre-trained text encoder from OpenCLIP 2 ViT-H/14 to obtain semantic embeddings of text descriptions.”
Rerender A Video: Zero-Shot Text-Guided Video-to-Video Translation (2023, methodology): “We use GMFlow [41] for optical flow estimation and compute the occlusion masks by forward-backward consistency check.”
MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images (2024, methodology): “To construct the cost volumes, we first extract multi-view image features with a CNN and Transformer architecture [45, 46].”
TAPIR: Tracking Any Point with per-frame Initialization and temporal Refinement (2023, unknown): “Optical Flow focuses on dense motion estimation between image pairs [20, 34]; methods can be divided into classical variational approaches [4, 5] and deep learning approaches [14, 21, 43, 55, 57, 66].”