/tf-VGG16

VGG-16 in tensorflow 1.15

Primary LanguagePython

Tensorflow 1.15 implementation of VGG-16.

Download Imagenet weights here.

Requirements:

  • Python 3.7
  • Tensorflow 1.15

Usage:

  • training:
python main.py
  • predict:
python predict.py

Notes:

  1. ./files_npy/ : folder contatining the following npy files
  • train_files.npy
  • trains_labels.npy
  • val_files.npy
  • val_labels.npy
  • test_files.npy
  • test_labels.npy

*_files.npy contains filenames of images randomly chosen for each set.

*_labels.npy contains integer labels for the images setted in *_files.npy (must be in the same order).

  1. ./data/ contains the images.
  2. ./tmp/ where the weights are saved during training.
  3. ./losses/ where the loss and accuracy metrics are saved during training.