Text this: Deep-learning-based multi-transducer photoacoustic tomography imaging without radius calibration