search

kjunelee / MetaOptNet

Meta-Learning with Differentiable Convex Optimization (CVPR 2019 Oral)
Apache License 2.0
529 stars 97 forks source link

Pretrained model #30

Closed hytseng0509 closed 5 years ago

hytseng0509 commented 5 years ago

Hi,

Thanks for sharing the source code! I'm wondering if you can provide the mode file you use to report the number in the paper. In particular, I would like to ask for the model file "MetaOptNet-SVM-trainval" evaluated on the 5-shot mini-ImageNet classification (which reports 80.00 ± 0.45%). Thanks!

kjunelee commented 5 years ago

Hi I was able to find a pertained model with 5-shot performance of 79.90 ± 0.44 % in my server. I just ran the code minutes ago and 1-shot performance of this pertained model is 63.75 ± 0.62 %.

Here is the link to the pertained model: https://drive.google.com/file/d/1NQHE1fCflrEDrBxz_6QURJUZ2JPWZZtJ/view?usp=sharing Hope this helps.

hytseng0509 commented 5 years ago

Thanks!