In this research, the three-dimensional structural and colorimetric modeling of three-dimensional woven fabrics was conducted for accurate color predictions. One-hundred forty single- and double-layered woven samples in a wide range of colors were produced. With the consideration of their three-dimensional structural parameters, three-dimensional color prediction models, K/S-, R-, and L*a*b*-based models, were developed through the optimization of previous two-dimensional models which have been reported to be the three most accurate models for single-layered woven structures. The accuracy of the new three-dimensional models was evaluated by calculating the color differences L*, C*, h degrees, and E-CMC(2:1) between the measured and the predicted colors of the samples, and then the error values were compared to those of the two-dimensional models. As a result, there has been an overall improvement in color predictions of all models with a decrease in E-CMC(2:1) from 10.30 to 5.25 units on average after the three-dimensional modeling.