OpenCV seems like a good platform for academics to publish their algorithm as code. It's good that to try a new approach I don't need to figure out how to get/c

OpenCV seems like a good platform for academics to publish their algorithm as code. It's good that to try a new approach I don't need to figure out how to get/compile/install a new tool. But we need something better for doing real work... something that simply offers the most basic building blocks. Easy GPU offloading, easy inputing/outputting of movies or images, cross platform shader tools, colorspace conversion, image filters, etc.