CRIPAC-DIG / NGNN

[WWW 2019] Code and dataset for "Dressing as a Whole: Outfit Compatibility Learning Based on Node-wise Graph Neural Networks"
80 stars 24 forks source link

can you provide more detailed guide to reproduce & predict (which files for what and provide the prediction files) if possible? #9

Open vyaslkv opened 5 years ago

AlexG1105 commented 4 years ago

Here is what you do:

  1. Run the train, and it will save the best checkpoint. As it is training, they calculate fitb and AUC score which you can see in the code and study,
  2. Load the graph, and then do the same thing with fitb and AUC on the data you want to test to produce the outfit scores.
surheaven commented 3 years ago

Here is what you do:

  1. Run the train, and it will save the best checkpoint. As it is training, they calculate fitb and AUC score which you can see in the code and study,
  2. Load the graph, and then do the same thing with fitb and AUC on the data you want to test to produce the outfit scores.

@AlexG1105 I have finished the first and I don't know how to "Load the graph, and then do the same thing with fitb and AUC on the data you want to test to produce the outfit scores". To get the second, which code can I run?

surheaven commented 3 years ago

Here is what you do:

  1. Run the train, and it will save the best checkpoint. As it is training, they calculate fitb and AUC score which you can see in the code and study,
  2. Load the graph, and then do the same thing with fitb and AUC on the data you want to test to produce the outfit scores.

how to load the graph?