2018 Update: Some more revisions in preparation for tackling other spinosaurids (hopefully!) in the coming year. I've decided that there is probably less missing from the incomplete femur (missing a section out of the middle) than before, which brings the femur more into line in terms of how robust it is relative to its length. I updated the scaling of the incomplete lower limb elements to reflect his. I also updated the vertebral column somewhat, reducing the rate at which the neural spines increase in height in the mid-posterior dorsals. Along this line, I've seen some internet hacks of my skeletal (note: please don't hack up or modify my skeletal reconstructions and post them!) that restore the last dorsal as having a very short neural spine; the neural spine of the last dorsal is obviously broken and is clearly labeled as such in the classic Charig & Milner description, so there's no reason to try and force it to have a really short neural spine. Finally, I've raised the neck a bit - not due to any changes in the anatomy, but to put it more in line with the semi-flexed pose I use on other theropods, and because I think it will make a more useful comparative pose in its relatives. For posterity I've left the previous updates below.
2016 Update: I updated the presacral series to reflect the new identifications of Evers et al. 2015 in their Sigilmassasaurus description. This has the effect of putting a stronger S-curve back into the neck, but it still leaves us with a hangdog angle for the skull. Interestingly, the neural spine morphology suggests the building up of axial muscles or nuchal ligaments (or both) along the back of the neck and front of the dorsal column, which is not unlike what Andre Cau has suggested for Spinosaurus, and would make it analogous to what we see in Deinocheirus as well (but on a smaller scale than either of those taxa).
2015 Updated: After nearly a decade and a half here is the overhauled skeletal. The overall proportions aren't all that different, but some of the details are. The midline crest has been moved back above the lacrimal, and I can now confirm that the odd downcurving neck seems to be a real thing, although it also uses some upwardly deflected almost cervicalized anterior dorsals to achieve it. The gray portion of the ilium is the part that was preserved as an imprint (i.e. there is no surviving bone from those parts) and so its accuracy depends entirely on the observations of the original excavators.
Original description: Baryonyx wasn't the first spinosaurid found, but it went a long ways in clarifying what these sorts of theropods looked like and ate. And it turns out they ate fish - although like modern crocodilians, Baryonyx almost certainly ate anything else it could get a hold of too (both fish scales and iguanodont bones were found in its stomach).