DeepBubbleVelocimetry model weights (DBV_MFFV)

Description Trained model weights for DeepBubbleVelocimetry (DBV), a deep learning-based optical flow method for measuring bubble velocities in two-phase bubbly flows (Choi & Kim et al., 2022, Sci. Rep.). The weights file (DBV_MFFV) contains a fine-tuned PWC-Net checkpoint (step 761,000) trained on synthetic bubble images. The network was initialized from a PWC-Net model pre-trained on FlyingChairs and FlyingThings3D, then fine-tuned on synthetic bubble image pairs generated with the included SyntheticBubbleImage generator. Contents of Weights.zip `Weights/DBV_MFFV.data-00000-of-00001` — Model weights (TensorFlow 1.x checkpoint) `Weights/DBV_MFFV.index` — Checkpoint index file `Weights/checkpoint` — Checkpoint pointer file Usage See the GitHub repository for inference instructions: https://github.com/dae416/DeepBubbleVelocimetry See the paper for details: https://doi.org/10.1038/s41598-022-16145-y