You should run commands in the src folder of the project directory. Delete the features.pkl, model_0.h5, and descriptions.txt files before running. These will be generated from preprocessing and training the model
Note that you can run the caption generation model with the two layered lstm, the gru, or the one layered lstm
Note: a pretrained vgg16 is used in the main project. An implementation of vg16 for 10 classes can be run from the root directory (vgg16.py). Download this dataset and rename folder to vgg16-data before running the file.
You can find the final writeup here.