Is there any available, ready-to-use model/nets?

[bontorhumala]

Hi Kelly, thanks for sharing your work. I am trying to run this code, but encounter the same problem as #2

I understand that you are not supporting this repo anymore. However, I am just curious, do you have trained nets that we can just use for your test set? Because I see each experiment code (e575, e576, etc) and I guess all of them require nets for each appliance to be created/trained from scratch. And I dont see trained nets inside neuralnilm subdirectory.

Thanks again!

[JackKelly]

Hi,

Thanks for your interest. When I'm next in the office, I'll check to see if I have any trained neural nets that I can provide. No promises, though!

Have a good weekend,
Thanks,
Jack

[JackKelly]

I've uploaded a few nets to my FTP server:

  host: achillea.doc.ic.ac.uk
  port: 55555
  username: anonymous
  password: <your email address>

[gohingsiang]

Hi, may i have the pretrained model as well? Thanks in advance.

[JackKelly]

thank you for your interest. Now that I have left Imperial, I'm afraid that pre-trained models are no longer available. Sorry.

I'd highly recommend using a modern DeepLearning framework like TensorFlow for NILM work now-a-days.

[gohingsiang]

Thanks for the reply, maybe i will just do it step by step. By the way, your advice was most appreciate.

[maechler]

For anyone interested in this, I have successfully reimplemented Jack Kelly's rectangles architecture using Keras and made it available here: https://github.com/maechler/nnilm

[JackKelly]

Great work! Thanks!