samwuko/siamese-gan-triple-network

This project solves the problem of cross domain image retrieval.

Network Architecture

Requirements

• Python
• NumPy
• Tensorflow>=1.4
• keras>=1.3

Experiment Dataset

• CVUSA datset: a dataset in America, with pairs of ground-level images and satellite images. All ground-level images are panoramic images.
  The dataset can be accessed from https://github.com/viibridges/crossnet

Usage

Step 1. Download this repository with git or click the button.

Step 2. Use input_data.py to extract the VGG features of the image, and all the features of the training set and test set will be saved in the data file.

$ python input_data.py

Step 3. Use the data_processing.py file to separate the VGG features under the data folder and save them to the data_vgg folder.

$ data_processing.py

Step 4. Training network and computing accuracy on test set.

$ python trian.py 

Note: The default parameters of batch size is 100, and epoch 100. You may need to modify the train.py file

Results

After each iteration, the accuracy of the model is calculated. Precision data is stored in accuracy.txt under model (automatic generation) folder.