Mid-infrared imaging detectors are essential tools for many applications because they can visualize the objects in the dark via thermal radiation. However, these detectors have to pair with separate spectral and polarization filters to select the target spectral bands and polarization states, resulting in complicated and bulky imaging systems. One way to mitigate the need for separate spectral filters and polarizers is to use metamaterial absorbers, which are arrays of optical resonators with sub-wavelength dimensions and spacing, to tailor the responses of the detector pixels. Here we report an intelligent program based on the genetic algorithm that automates the design and optimization of a metal-insulator-metal based metamaterial absorber with multi-sized nanostrip antennas as the top layer. The program starts from a randomly generated pattern of the top antenna layer, and it iteratively approaches the optimized designs of two polarization selective MIM absorbers with wideband high absorption in the specified 3-5 (MWIR) band and 8-12 µm (LWIR) band. The measured absorption spectra of the two optimized designs agree well with the simulated results. The influences of the incident angle of light, the finite size of detector pixels, and the air gap between the neighboring pixels on the spectral absorption are numerically evaluated.