Recommended DNN Framework [closed]

asked 2019-02-14 04:41:31 -0600

mbeer gravatar image

updated 2019-02-14 05:49:22 -0600

Hi,

want to use OpenCV's DNN module for several tasks in several projects: classification, detection, and segmentation. I know there are pre-trained models which I can download, but i will have to re-train for my application.

I have tried a few examples from the model zoo and they worked.

I would now like to train a network from scratch or transfer train with my own classes, and load&run the resulting model in OpenCV.

I have tried this in Tensorflow but didnt succeed. I have failed so far to generate pb (and pbtxt)-Files which could be read by OpenCV. I followed tutorials for training and transfer training, preparation of model Files, etc. But overal it was a rather un-satisfying experience.

My impression is that Tensorflow is great for researchers, and my focus is on application. At least OpenCV is also about applying trained DNNs. Now my question is: If I want to use an existing network configuration, just (transfer-)train it with my own data, and apply the model in OpenCV, should I use a different DNN Framework from Tensorflow? And which one?

Much appreciated, thanks!

edit retag flag offensive reopen merge delete

Closed for the following reason question is off-topic or not relevant by mbeer
close date 2019-02-21 08:03:26.702634

Comments

2

off topics. Use what you want. If you got a problem with your network in opencv, ask a question here and opencv team will try to solve it

LBerger gravatar imageLBerger ( 2019-02-14 04:48:13 -0600 )edit
1

you won't believe it, but it's unclear, what you're asking.

recommended framework for what ?

opencv's dnn does inference (aka the "forward pass" only) on pretrained models, and there's no extra framework needeed for that, so, what is your question, again ?

berak gravatar imageberak ( 2019-02-14 04:50:30 -0600 )edit
1

Thanks for your comments, I will re-write the question! Sorry for the ugly start, i hope I can be clearer!

mbeer gravatar imagembeer ( 2019-02-14 05:12:13 -0600 )edit
1

^^ sure, please, you're welcome, nonetheless, heeh ;)

berak gravatar imageberak ( 2019-02-14 05:15:58 -0600 )edit

i don't think, there can be a general answer here.

to my exp. usually, you have a pretrained model, you want to work with / improve, and you have to use the framework / code, it was trained with

berak gravatar imageberak ( 2019-02-16 02:38:27 -0600 )edit