<p>Flow metering is an essential industrial requirement commonly provided by differential pressure flow meters that have an obstruction device, causing huge upstream pressure at higher flow rates. Thus, the present study measures the flow velocity based on the pressure drop in the straight pipeline without any restriction device. Since the physical equations cannot be explicitly solved for the velocity calculation from pressure drop data, and the Trial-and-Error (TAE) procedure is time-consuming and unreliable, the deep learning approach is employed to determine the flow velocity from pressure drop data. Darcy friction factor and pressure drop are calculated to a wide range of flow velocity, pipe diameter, pipe relative roughness, and fluid kinematic viscosity to generate more than 27,000 theoretical data for training and testing different deep neural networks such as FFNN (Feed Forward Neural Network), CNN (Convolutional Neural Network), LSTM (Long and Short-Term Memory), and RNN (Recurrent Neural Network). Statistical analysis proves the influence of input variables and justifies their selection. Models have been evaluated via experimental data that include large pipe diameters up to 10cm and different fluids such as water and heavy oils with a kinematic viscosity of 340 m²/s. The FFNN with a 4-15-5-5-1 structure and the CNN with four layers had the best response, as their accuracy is more than 96% and 95% for all data, respectively. This study presents a new, reliable, and inexpensive flow metering system for incompressible fluids with a low computational cost, no additional mechanical parts, and no intrusive hardware.</p>