There are lots of freely available source code for pattern recognition and machine learning algorithms. SourceForge not only has code and whatnot, but various projects such as OpenPR which are dedicated to machine learning, including specifically pattern recognition (UCI's machine learning repository has a ton of categorized datasets). But apart from working, what makes any particular code "good" is largely what you want to do with it. There are certain CONDITIONAL exceptions, such as toolboxes for MATLAB, statistical/scientific computing software environments with built-in functions, and in some cases source code designed to be generic (e.g., a class, library, etc.; often intro books on programming have these, e.g., The Nonlinear Workbook: Chaos, Fractals, Cellular Automata, Neural Networks, Genetic Algorithms, Gene Expression Programming, Support Vector Machine, Wavelets, Hidden Markov Models, Fuzzy Logic with C++, Java and SymbolicC++ Programs, code here: http://issc.uj.ac.za/NonlinearWorkbookCode/4editioncode.zip).
However, if you are asking for specific code and whether it was "good", I'd say "yes" and "no" to every single code I've written, used, evaluated, or seen run. That's because if I want my pattern recognition code to run through a rather badly constructed database of information about documentation for a particular kind of tax code and generate classes/categories using various (in this case mostly supervised) classification algorithms I can use to create a program that answers queries about what kind of documentation, request forms for documentation, etc., is needed for X particular action/process by associating the queries with generated requests corresponding to probable classes/categories generated by the pattern recognition code which my client can use to use to then narrow results by further input that is then matched again with already generated classes/categories until the user finds the correct match or matches, any source code that using cutting-edge, highly sophisticated pattern recognition to analyze CAPTCHAs is worthless.
Granted, most of the code I've found or looked for was for teaching purposes or to help me work with a language I was not or am not very familiar with, but in my experience a lot of the source code freely available requires so much work to be adapted to one's specific needs (or is designed to be so adapted), that unless this is what one is looking for there isn't much point in finding free source code. As I say, though, I've never actually spent time looking for code for a particular pattern recognition application, but I'd say the problems follow simply given how many different kinds of algorithms can be counted as pattern recognition algorithms or methods/techniques (ANNs, fuzzy ANNs, k-means clustering, SVMs, dimensionality reduction & feature extraction methods, decision trees, hidden markov models, not to mention things like corpus linguistics and prepared ontologies, n-gram datasets, etc.).
Much easier to answer question! Alas, I haven't used FORTRAN since my father (who started programming with punch cards) started teaching us basic programming. Due to this psychological trauma (I jest), I haven't used FORTRAN since. I also haven't come across many examples of pattern recognition or related algorithms written in FORTRAN ever. Python, Java, C, C++, C#, MATLAB, R, S-Plus, SAS, and others I am less or not familiar with, sure. FORTRAN not so much. But
1) Knowing Fortran makes learning better languages easier.
&
2) Others may know of Fortran algorithms/code that is of the kind you are looking for.