leonidmedved/TOFD-dataset: TOFD-dataset

This dataset is intended for training and validating neural networks for defect segmentation from TOFD (Time‑of‑Flight Diffraction) data.

Includes:

Synthetic data — echo signals simulated in the CIVA software package using the pencil method.
Real data — measurements performed on blocks with purposely introduced defects.

Data are split into training and validation sets. The dataset can be used for:

training segmentation models (U‑Net, U‑Net++, Attention U‑Net, etc.),
testing robustness on synthetic vs. real data,
developing signal filtering and post-processing methods.