[DNN] Clarify the model deployment approach with OpenCV stakeholders

[huningxin]

No description provided.

[huningxin]

@akineeic , could you please host the current examples in your personal github and share the link? I can use that to demo and communicate with OpenCV team. Thanks!

[akineeic]

I set github page for image classification example (picture and camera). But the file loading speed of the webpage is very slow since it needs several minutes to load the bvlc_googlenet caffemodel (53MB).

[huningxin]

Looks good. Do you hosted multiple models? So user may try other models by change the URL?

[akineeic]

Not yet. But user can change the URL according the JSON file in this repo. The URLs for all ONNX models, squeezeNet caffe model and inception tensorflow model are availble.

[huningxin]

For GoogleNet in the tutorial, why you don't use ?
"modelUrl": "http://dl.caffe.berkeleyvision.org/bvlc_googlenet.caffemodel",
"configUrl": "https://raw.githubusercontent.com/BVLC/caffe/master/models/bvlc_googlenet/deploy.prototxt"

[akineeic]

For GoogleNet in the tutorial, why you don't use ?
"modelUrl": "http://dl.caffe.berkeleyvision.org/bvlc_googlenet.caffemodel",
"configUrl": "https://raw.githubusercontent.com/BVLC/caffe/master/models/bvlc_googlenet/deploy.prototxt"

Because the request of caffemodel will be blocked when using this modelUrl.