In industrial processes, soft sensor models are commonly developed to estimate values of quality-relevant variables in real time. In order to take advantage of the correlations between process variables, two convolutional neural network (CNN)-based soft sensor models are developed in this work. By making use of the unique architecture of CNN, the first model is capable of utilizing abundant process data, and the complexity of this model remains low. The second model integrates finite impulse response and CNN, and process dynamics can be reasonably embraced in this model. The effectiveness of the two models is validated by a simulation case and a chemical industrial case. Since the finite impulse response-convolutional neural network (FIR-CNN) can give the best prediction accuracy and the most interpretable trend of a quality-relevant variable, it has promising application potential in the chemical industry.