znhu / image-caption-using-vgg-lstm-attention-model Goto Github PK

This is a VGG-LSTM based End-to-End Model with attention, for image caption

Python version support:

Python 3.7

Python packages satisfied:

PyQt5, pickle, pycocotools, cython, nltk, gensim, PIL, numpy, tensorflow

To be noticed, version of tensorflow == 2.1.0

To run the system, ensure the folowing document are in the GUI package:

show3.7.py deal.py Dic.txt dictionary wordvector_dictionary background.jpeg Model_cat.hdf5 Model_decoder.hdf5 Model_decoder.hdf5

How to run the system:

step 1: open terminal step 2: type cd ../GUI on terminal step 3: type python show3.7.py on terminal

Python files' content:

deal.py : Model building modules

show3.7.py : Interation board modules

To satisfy the above running, you should download the relative code with path below, and put them in the same load with the code files.

https://drive.google.com/file/d/137IEP5DxMa6-AkY9bh2ydQCm4NT06DGc/view?usp=sharing

If you prefer to get a new model with your own setting, you could get the dataset from the path below:

Dataset resourse:

https://cocodataset.org/#overview

References:

LSTM Decoder by Keras: https://keras.io/examples/lstm_seq2seq/ Attention layers by Keras: https://www.tensorflow.org/api_docs/python/tf/keras/layers/Attention