Added in faster rcnn

[benhoff]

No description provided.

[nmanovic]

@benhoff , the script also doesn't work.

[benhoff]

@nmanovic Which version of openvino are you using? There's an API change in openvino 2019 that I've accounted for locally that I'm not sure how to add into the project

[nmanovic]

@benhoff, I'm using the latest version of OpenVINO 2019 R1.1 (144).

[benhoff]

@nmanovic see openvinotoolkit/openvino#128

Specifically, alalek's comment on April 25th:

MO 2018R5 generates model IRs with these inputs/outputs (see the first post about used parameters of MO):

network.inputs = ['image_tensor']
network.outputs = ['detection_output']

MO 2019R1 IRs:

network.inputs = ['image_tensor', 'image_info']
network.outputs = ['detection_output']

If you look at the Cvat line that handles the input for OpenVino, it's pretty simple.

Relevant code: https://github.com/opencv/cvat/blob/021b77cceeab5927632684b737cd11f66ce087dc/cvat/apps/auto_annotation/model_loader.py#L31

I've hacked in some handling locally. I can split it out of my local dev and drop a pull request and we can talk about the best handling from there?

[benhoff]

See: #545