Is it possible to translate OpenVino person-detection-retail model to another framework, like tensorflow? about open_model_zoo

[snosov1]

Is it legal to translate it to TesorFlow and use it in a commercial product?

Sorry, I don't think you can find any legal advise here in the open. I suggest to consult the legal department in your organization (that will, supposedly, produce the "commercial product"). The repo is licensed under Apache 2.0.

We start releasing some of the models in the original format (see https://github.com/opencv/training_toolbox_tensorflow and https://github.com/opencv/training_toolbox_caffe for example). It might better fit your needs.

Maybe it is just better to train YOLO model for one class person?

I would strongly advise against training YOLO =) In our experience, SSD-like architectures provide a much better accuracy/peformance trade-off.