What do you like best?
We've used OpenCV on a handful of projects. For each one, we've been able to accomplish a difficult Computer Vision (CV) task with relatively little engineering effort. This doesn't mean it hasn't taken any work, but that it's taken significantly less work and expertise than building a custom solution. Using OpenCV with an adapter like Cylon to a high-level language allows for rapid prototyping of ideas that include or rely on CV.
What do you dislike?
OpenCV, when used with its native interface, is often difficult to implement against. You need a developer knowledgeable about CV and that also knows either C or C++. This is not easy to find in the current job market. In general, OpenCV paired with adapters to higher level languages alleviate this need. Particularly on the server side, working in node.js is very natural and allows the front and backend stack to stay in the same language.
Recommendations to others considering the product
I'm definitely a proponent of OpenCV. I've used it in multiple projects and despite not having a strong background in CV, it's met my needs and allowed us to ship products using it as the technology underpinnings.
What business problems are you solving with the product? What benefits have you realized?
Rapid prototyping of a human-digital signage platform, feature extraction for educational document scaning and automated grading, computer-human interaction engine for a smart cooking assistant, as well as several other CV-based features. We've implemented on the desktop and server-side in addition to implementations on more memory confined platforms like mobile devices (iPad/iPhone).