dblalock / imagenet_validation

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How to reproduce ImageNet validation results

This code in this repository can be used to reproduce the ImageNet validation results for Keras pretrained models. A blog post describing this process in more detail is here.

Setup instructions

The ILSVRC2012 validation images should be put in data/val/. The remaining files mentioned in the blog post, and enumerated below, should be put in data/:

  • ILSVRC2012_validation_ground_truth.txt
  • meta.mat
  • synset_words.txt

After the data is organized the notebooks can be run in order to reproduce model validation results.

About


Languages

Language:Jupyter Notebook 100.0%