Amongst the main issues in extra-galactic astrophysics that can be addressed by ISO are: (1) the relationship between starburst galaxies and AGN; and (2) the ability of AGN unification schemes to explain the FIR emission of AGN. To do this requires the photometric study of a large, statistically complete sample. We propose measuring the FIR SEDs of a complete, cleanly-defined sample of IRAS galaxies selected on the basis of their 25 micron flux. The FIR is surpassed only by the X-ray region in its ability to select unbiased samples of AGN, and, by selecting galaxies about a 25 micron flux cut, we shall obtain a much higher fraction of starbursts and AGN than is found in the many other samples of IRAS galaxies for which photometric follow-up projects are being undertaken with ISO. Our sample will include some normal galaxies, which will be used as a control sample for understanding the far-infrared emission processes of active galaxies: their presence is unavoidable, since IRAS colour information is insufficient by itself for selecting only starbursts and AGN, which is the reason that we require the greatly increased quality of photometric information obtainable from ISO. The library of far-infrared SEDs produced by this project will complement those obtained by other ISO galaxy samples, which typically select a much higher fraction of normal galaxies, to build up a picture of the full range of far-infrared emission from galaxies. These data will fuel ongoing research at Imperial College into techniques for modelling far-infrared emission. The improved far-infrared models and observed SEDs that will result from this project, will enable us greatly to advance the study of the AGN and starburst phenonmenae in galaxies, in particular testing AGN unification schemes.