This project is nowhere worthy of being called a "biometric SDK". It looks like a college student decided to post his class project from a "Computer 'Vision'" (or equivalent) class on his research into fingerprint analysis.
What IS inside the so-called "SDK" is:
- A bunch of functions that take a grayscale image as input and alter the image in an attempt to find minutiae. None of the functions are exported (i.e. NO DLL). On top of that, the techniques used seem fairly "hacky" - if you know what I mean.
- Requires pre-scanned images as input (several are included as examples). That is, there are NO scanner hooks. This means you still have to write code to integrate with another SDK specifically designed for the scanner.
- There are Java and C# (NO C++) implementations of the same code. Apparently the Java version has some dependency on Eclipse (see other forum topics). Tried the C# version and it "sort of" worked.
- There is ZERO documentation on this project. And apparently it doesn't work for most people as a result.
- This is labeled 'Alpha'. It should be 'Pre-Alpha' because this is more of a 'Demo-that-sucks'. 'Alpha' should be reserved for when you actually have something working that people who are the "bleeding-edge" types can try out.
Those who are looking for a solution that integrates multiple scanners with a common identification solution from a single SDK are going to have to hunt for commercial solutions at this point (I too was hoping there was something to this SDK before I shelled out big bucks). My guess is that this is a dead project (started in the middle of August of 2005 and doesn't have much to show for it today [March 23, 2006]).
Open source dose not mean it is free production level software .It is a way of building prototype software for the benefit of the community.
You think it could be better make some improvements and submit the changes. The image processing needs some work so if you any good at this hopefully you would be able to improve this.
A REAL Biometric SDK integrates with at least one scanner device attached to the computer. Your code takes pre-scanned images and uses a number of very weak algorithms. My complaint is that you call this 'Alpha' quality software...which means it is ready for bleeding-edge types and implicitly has support for at least one scanner _natively_. This project should not be labeled anything higher than 'Pre-Alpha'.
I'm not saying it has to be "production ready" software, just don't lie to people and say it is something it is not. All I'm doing here is letting people know that if they are looking for a Biometric SDK, they need to look elsewhere for the time being as this isn't anywhere near ready for use by even the most cutting-edge developers. I've carefully analyzed your source code and it looks like a college student's attempt at basic recognition. Hardly worthy of a sourceforge project at this point.
You should also be aware that a search for "Biometric SDK" on Google turns up this project as the _FIRST_ link. Expectations of visitors, therefore, are that the project is up to par with said search result position. That is, they think it should be production-ready and should be able to hold its ground when compared to commercial packages. It isn't and it doesn't. This is why people are confused when they download the package and it doesn't work.
Frankly, this project is being developed incorrectly. You need multiple components:
- Native scanner image acquisition
- Low level minutiae extraction with rotational scan support
- High level database loading, storage, and comparisons
- High level functions to take an image and perform 1:1 and 1:N matching/verification against the database.
- A method of accessing above components from multiple languages (e.g. a DLL, possibly COM-based).
- Plenty of documentation on each exported function in the DLL complete with examples.
- Later on add MSGINA replacements.
The current project is just a bunch of hacked together algorithms that barely function. I recommend scrapping the current project and starting again using a more sensible approach that other developers like myself would be interested in. Your current approach won't get anyone to help you.
I recommend restarting the project this time focusing first on image acquisition from, say, a half-dozen scanners. Pick a platform (e.g. Linux or Windows), drop the Java crap (use something that can make shared libraries/DLLs), and start work on the core shared library for acquiring images from fingerprint scanners. Once that is stable, then you can start work on the second item. Once the second item is working pretty well, then people who actually know how to write code will show up to help...but not before that point. It is a sad but true fact.
Thank you syphertext
Log in to post a comment.
Sign up for the SourceForge newsletter:
You seem to have CSS turned off.
Please don't fill out this field.