In circular scanning photoacoustic tomography (PAT), it takes several minutes to generate an image of acceptable quality, especially with a single element ultrasound transducer (UST) as detectors. Although, imaging speed can be enhanced by employing faster scanning (with high repetition rate light sources) and using multiple USTs, it is still limited due to artifacts arising from the sparse signal acquisition and low signal-to-noise (SNR) at higher scanning speeds. Herein we propose a novel method to improve the framerate (or imaging speed) of circular PAT systems by utilizing the potential of deep learning. We developed a U-Net-based convolutional neural network termed as hybrid dense UNet (HD-UNet) to reconstruct the PAT images from fast-scanning acquired data. The efficiency of the proposed network was evaluated on both single- and multiple-UST based PAT systems. Our results on both phantom and in vivo imaging demonstrate that the proposed network can improve the imaging frame rate by ~6 fold in single-UST based PAT systems and by ~2 fold in multi-UST based PAT systems without compromising the quality and structural integrity of the image. We achieved the fastest frame rate of ~3 Hz imaging with multi-UST system. This is the fastest reported imaging speed so far in the literature based on single element transducer scanning PAT.