Rotating drum experiments on binary mixtures of plastic spheres with wood spheres, wood cylinders, plastic cylinders, and wood cubes were investigated, respectively. A high-resolution camera was employed to record the flow behaviors of the binary mixtures. A machine learning-assisted image processing method was developed to segment the particles of different shapes, and its superiority was demonstrated by comparing the results with the pixel and artificial counting methods. This method was used to analyze the images recorded from binary mixture experiments to explore the effects of rotating speed, particle shape, and density differences on mixtures' radial and axial mixing and segregation behavior. The results show that the particle shape and density difference significantly affect the resolution angle of the binary mixture containing plastic spheres. In the radial direction, the degree of mixing of the binary mixtures containing wood spheres and cubes is not sensitive to speed. The degree of mixing of the binary mixture containing cylindrical particles increases linearly with the rotating speed within a specific range. The particle shape significantly affects the radial segregation of the binary mixtures. In contrast, the rotating speed has a negligible effect on the radial segregation of binary mixtures containing wood spheres, cylinders, and cubes.