Learning to Estimate and Refine Fluid Motion with Physical Dynamics

Authors: Mingrui Zhang, Jianhong Wang, James B Tlhomole, Matthew Piggott

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed approach outperforms optical flow methods and shows competitive results compared to existing supervised learning based methods on a benchmark dataset. Furthermore, the proposed approach can generalize to complex real-world fluid scenarios where ground truth information is effectively unknowable. Finally, experiments demonstrate that the physical corrector can refine flow estimates by mimicking the operator splitting method commonly utilised in fluid dynamical simulation.
Researcher Affiliation Academia 1Department of Earth Science and Engineering, Imperial College London, UK 2Department of Electrical and Electronic Engineering, Imperial College London, UK.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code is available at: https://github.com/erizmr/Learn-to-Estimate-Fluid-Motion.
Open Datasets Yes The PIV dataset is a synthetic dataset collected by (Cai et al., 2019b).
Dataset Splits No The predictor is trained on the PIV dataset with 12,190 samples used for training and 1505 for testing. A specific validation split size or methodology is not explicitly mentioned.
Hardware Specification Yes All experiments are conducted on a moderate level GPU Nvidia Tesla P100 16GB.
Software Dependencies No The paper mentions using the Adam optimiser but does not specify version numbers for any software libraries, frameworks, or programming languages used.
Experiment Setup Yes We train the model for 40,000 iterations with a batch size of four image pairs using the Adam optimiser. The learning rate is kept at 10 4.