In the name of God
This repository contains implementation of ZFNet (Visualizing and Understanding Convolutional Networks) by Tensorflow and the network tested with the CIFAR-10 dataset.
Before train and evaluate the network, you should download the following dataset:
- CIFAR-10 Dataset: https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
Extract the cifar-10-python.tar.gz
file, then your folder structure should be like the following image:
To train the network with cifar-10 dataset, type the following command at the command prompt:
python3 ./train.py
Sample images from cifar-10 dataset:
Train Accuracy = 0.100
Test Accuracy = 0.100
Train Accuracy = 0.215
Test Accuracy = 0.216
Train Accuracy = 0.364
Test Accuracy = 0.357
...
Train Accuracy = 0.994
Test Accuracy = 0.728
...
Final Train Accuracy = 1.000
Final Test Accuracy = 0.753
To evaluate the network with cifar-10 dataset, type the following command at the command prompt:
python3 ./evaluate.py
- Python 3
- numpy
- scipy
- pillow
- tensorflow