Abstract:The accuracy of learning-based optical flow estimation models heavily relies on the realism of the training datasets. Current approaches for generating such datasets either employ synthetic data or generate images with limited realism. However, the domain gap of these data with real-world scenes constrains the generalization of the trained model to real-world applications. To address this issue, we investigate generating realistic optical flow datasets from real-world images. Firstly, to generate highly realistic new images, we construct a layered depth representation, known as multiplane images (MPI), from single-view images. This allows us to generate novel view images that are highly realistic. To generate optical flow maps that correspond accurately to the new image, we calculate the optical flows of each plane using the camera matrix and plane depths. We then project these layered optical flows into the output optical flow map with volume rendering. Secondly, to ensure the realism of motion, we present an independent object motion module that can separate the camera and dynamic object motion in MPI. This module addresses the deficiency in MPI-based single-view methods, where optical flow is generated only by camera motion and does not account for any object movement. We additionally devise a depth-aware inpainting module to merge new images with dynamic objects and address unnatural motion occlusions. We show the superior performance of our method through extensive experiments on real-world datasets. Moreover, our approach achieves state-of-the-art performance in both unsupervised and supervised training of learning-based models. The code will be made publicly available at: \url{<a class="link-external link-https" href="https://github.com/Sharpiless/MPI-Flow" rel="external noopener nofollow">this https URL</a>}.

OIFlow: Occlusion-Inpainting Optical Flow Estimation by Unsupervised Learning

Unsupervised Learning of Scene Flow Estimation Fusing with Local Rigidity.

A Dense Optical Flow Registration Algorithm Based on Deep Learning

NccFlow: Unsupervised Learning of Optical Flow With Non-occlusion from Geometry

Learning By Analogy: Reliable Supervision From Transformations For Unsupervised Optical Flow Estimation

OAS-Net: Occlusion Aware Sampling Network for Accurate Optical Flow

SplatFlow: Learning Multi-frame Optical Flow via Splatting

Unsupervised Learning of Optical Flow with Patch Consistency and Occlusion Estimation.

Self-Attention-Based Multiscale Feature Learning Optical Flow with Occlusion Feature Map Prediction

STFlow: Self-Taught Optical Flow Estimation Using Pseudo Labels

Unsupervised Learning Optical Flow in Multi-frame Dynamic Environment Using Temporal Dynamic Modeling

Optical Flow Estimation Via Motion Feature Recovery

OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation

MPI-Flow: Learning Realistic Optical Flow with Multiplane Images

StreamFlow: Streamlined Multi-Frame Optical Flow Estimation for Video Sequences

SKFlow: Learning Optical Flow with Super Kernels

MRDFlow: Unsupervised Optical Flow Estimation Network With Multi-Scale Recurrent Decoder

Unsupervised Learning of Facial Optical Flow via Occlusion-Aware Global-Local Matching.

An Unsupervised Optical Flow Estimation For LiDAR Image Sequences

CLIP-FLOW: CONTRASTIVE LEARNING WITH ITERATIVE PSEUDO LABELING FOR OPTICAL FLOW