search

mmmmimic / OOD-CV-challenge

1 stars 0 forks source link

Did you train your models on ImageNet-1K from scratch? #1

Closed khawar-islam closed 1 year ago

khawar-islam commented 1 year ago

Dear @mmmmimic

Thank you for the work. Did you train Convnext and Swin ViTs from scratch on imageNet-1k? Did you use below dataset? https://www.kaggle.com/competitions/imagenet-object-localization-challenge/data

Regards, Khawar

mmmmimic commented 1 year ago

Dear Khawar,

Thank you for your interest in our work! We did not train ConvNext or Swin from scratch - we simply loaded the pre-trained parameters provided by pytorch-image-models. Here is an example from the official documentation.

BR, Manxi

mmmmimic commented 1 year ago

image

khawar-islam commented 1 year ago

Dear Manxi,

Thank you for your clarification.