A 3D WUCT system using a single ultrasound transducer is designed and automated. The dataset consist of the WUCT reconstruction results dataset used to train U-Net based semantic segmentation model.  Also, dataset i) to study the effect of increase in the number of virtual transducer on reconstruction quality and, ii) effect of variation in the applied pulse width on the reconstruction are provided. The U-Net based semantic segmentation model is trained and used to evaluate dice coefficient corresponding to the phantom’s actual profile and reconstructed profile.