/iMat-Product

iMaterialist Challenge (FGVC6, 2019), 13th place

Primary LanguagePython

iMat-Product

iMaterialist Challenge (FGVC6, 2019), 13th place

iMaterialist Challenge on Product Recognition (FGVC6, CVPR 2019): kaggle page

Backbone

I use the following 5 CNN backbones pretrained on ImageNet. All pretrained models are from pretrained-models.pytorch.

  • NASNet-A-Large
  • SENet154
  • InceptionResNet-v2
  • ResNet152
  • ResNet101

Settings

  • Learning rate: 4e-4
  • Weight decay: 5e-5
  • Epoch: all models converge in around 10 epochs

Fusing

Nothing special, average of all probabilities.