Predicting the center of MSCOCO images given the border and the captions
This code is a model written in tensorflow to predict the 32x32 center of a 64x64 image given the border and some captions.
Installation
pip install -r requirements.txt
Dataset
I have used the MSCOCO dataset and preprocessed the data into tf.Examples files. You can get my preprocessed dataset here.
mkdir dataset
cd dataset
tar xvf dataset_ift6366_images_captions.tar.gz .
To run the model:
python3 main.py
You can start a tensorboard instance to monitor the training:
tensorboard --logdir=logs
architecture
The architecture of the generator is resumed in this schema: 