"Deep Neural Networks Module: $150k" seems kind of unecessary. Torch/tensorflow already fills that niche so well.
Why would opencv, the all singing all dancing C++ computer vision framework not have a DNN module?
Why would it say “use a Python library”? (And yes, torch is deprecating its c++ support).
Not having it would be strange.
Here’s a starting point: https://discuss.pytorch.org/t/pytorch-2-and-the-c-interface/...
https://docs.opencv.org/4.x/d2/d58/tutorial_table_of_content...
In particular, OpenCV DNN is as far as I know the fastest CPU implementation for darknet/yolo networks.
At today's salaries, that pays for, what, 1 month of some ML engineers time to integrate all the big ML frameworks into OpenCV?
Doubt you could do any decent job of it in a month...