I ask opencv community, do they have a will to support PyTorch?Following is the answer
Quote from the discussion of #9645
The best drivers to start a research are interesting models trained in PyTorch. I think we don’t need a one more AlexNet represented in different format. It’ll be great if you can find or train some efficient network that outperforms known models from other frameworks. In example, any-class detection / tracking models, text detectors, background subtraction, style transfer.
Do you know any models worth to recommend? Thanks