Hacker News new | past | comments | ask | show | jobs | submit login
Removing Image Artifacts in C# with OpenCV (lostindetails.com)
42 points by jerleth 62 days ago | hide | past | web | favorite | 5 comments

Not that the C# isn't important here, but it's getting more and more important to be able to handle computer vision related tasks as part of a solution. Badge scanning, face scanning, orientation detection, etc... It's no longer feasible to write software that is a front end to a database (the twitters and the messaging apps of the world). We need to go beyond that and have something really capable. If Google Wallet now requests you to just place your credit card under the phone's camera - then that's the minimum expectation you're going to need going in front of VCs. They want that. Even if it's small - something way cool, cuts the competition out immediately. Cool sells.

Yeah, OpenCV is a C++ library with bindings for Python, Java, Rust, C#, etc. C# isn't really the important bit here.

OpenCV is a neat library though. I used it to do some quick and dirty measurements + sprite generation on a couple of hundred shark teeth for a group project in school. We made an educational virtual fossil dig website.


Does anybody know how good and up to date the C# wrapper is? I often find that the C# bindings are second or third class citizens in open source libraries that are either buggy or way behind. Saw this with Lucene and quite a few others.

Article Author here, you are totally on point: The wrapper is done by one contributor mainly https://github.com/shimat/opencvsharp/graphs/contributors

It works, but has it's problem, like memory leaks if you forget to dispose the objects and an upgrade to a newer version had a bug in it's ubuntu binaries that any simple test would have caught.

There is a higher quality commercial wrapper that is only free for gpl'ed solutions.

Which is what I see with some projects, even though the software itself is free, the c# wrapper done by a third party costs money which makes me loathe to use it, as it hinders reuse.

Take care, Martin

There is also Accord.NET, which is .NET-native and offers a lot of other machine-learning features beyond just image-manipulation: http://accord-framework.net/

I’ve used both off and on for the last 5 years and generally find Accord.NET to be more pleasant, though most of the attention online goes to OpenCV, so there tends to be more examples for it.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact