How to train the datasets?

[samylee]

Hi @stupiding ,
Thank you very much for sharing the code. Could you tell me how to train the datasets of AFLW,
and how to demo the caffemodels like 'model.caffemodel' you provided in the code?
Any suggestions will be great helpful to me!
Thanks!

Samylee

[stupiding]

I've uploaded the data list at face_detection_models/data. You can just checkout and get the corresponding images, then train the model with "caffe train" but make sure pycaffe have already been made. As for the demo, build/tools/FaceLib.bin can be used with along the test/param.cfg file in any of the model folders.

And unfortunately the scale 48x48 submodel is not well trained for now, I can only guess that the data imbalance brought in by the previous 12x12 and 24x24 submodel should be handled more sophisticatedly.

[samylee]

Hi @stupiding ,
Thank you for your quick reply.
I couldn't find the 'data list' at face_detection_models/data when I check it out, could you please tell me where it is?
And for demo, I don't know which argv[?] I should put behind the "build/tools/FaceLib.bin" when I look over the test/param.cfg, so could you please tell me how to use the FaceLib.bin? and the command line is better, like "FaceLib.bin argv[1] argv[2] argv[3] ...(Forgive me, I don't know how much the argumentation behind the FaceLib.bin)".
Sorry to bother again!
Thank you very much, I am looking forward to your reply.

Samylee

[stupiding]

The repository has been updated, you can checkout it again. And the usage of FaceLib can be seen with no parameters, then you can see something like this, Usage: FaceLib.bin <image_root_folder> <test_list_file><params.cfg>.