shampoodeng / raft-ss-piv-dev

RAFT with diffeomorphism integrated for PIV measurement.

RAFT-SS-PIV

todo

Requirement

All codes are running with python3.9, torch2.0.0

# For reading TFRecord
pip install tfrecord protobuf=3.20.1
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist --upgrade nvidia-dali-cuda110

conda install numpy matplotlib scipy jupyter
#conda install -c conda-forge openpiv

# For UnLiteFlowNet-PIV
conda install -c conda-forge cupy

Composition of datasets

Datasets for training and validation

Dataset for training model are avaiablie on https://doi.org/10.5281/zenodo.4432496. This dataset is stored in the format of TFRecord, whose basic elements are dictionary. Each data element has 3 sub-members: target, flow, label.

• target: particle image pairs, 1x2xWxH
• flow: flow field ground truth, 1x2xWxH
• label: dumped

Note: go to testTfrecord.py for more information

Datasets for evalution

Dataset for evaluating models is comprised of groups of sin flow and lamb-oseen flow with different parameter set. For both sin and lamb-oseen, each parameter set contains 10 samples. Similar to training dataset, evaluating dataset is in the form of dictionary, which has 4 keys: img1, img2, u, v.

• img1, particle image one, Bx1xWxH
• img2, particle image two, Bx1xWxH
• u, x componet of the flow ground truth, Bx1xWxH
• v, y componet of the flow ground truth, Bx1xWxH