https://github.com/tanyachutani/image-captioning-generator

Image Captioning Generator Keras
https://github.com/tanyachutani/image-captioning-generator

beam-search bleu-score caption-generation cnn cnn-rnn flicker8k-dataset greedy-search image-captioning inceptionv3 keras rnn tensorflow

Last synced: 28 days ago
JSON representation

Image Captioning Generator Keras

• Host: GitHub
• URL: https://github.com/tanyachutani/image-captioning-generator
• Owner: TanyaChutani
• Created: 2020-04-12T23:35:52.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2020-04-30T03:57:47.000Z (almost 5 years ago)
• Last Synced: 2024-11-11T21:26:26.707Z (3 months ago)
• Topics: beam-search, bleu-score, caption-generation, cnn, cnn-rnn, flicker8k-dataset, greedy-search, image-captioning, inceptionv3, keras, rnn, tensorflow
• Language: Jupyter Notebook
• Size: 156 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# ImageCaptionGenerator

Image Captioning Generator Keras

## Data

Dataset - Flickr 8k Dataset  


 


[Flicker8k_Dataset](https://github.com/jbrownlee/Datasets/releases/download/Flickr8k/Flickr8k_Dataset.zip) 


[Flickr8k_text](https://github.com/jbrownlee/Datasets/releases/download/Flickr8k/Flickr8k_text.zip)  


 


Flicker8k_Dataset - Contains 8092 images in jpeg format.  


Flickr8k_text - Each image contains 5 description. 


## Model

Built and trained a deep learning model for captioning real world image.

- Used pre-trained InceptionV3 to extract feature from image.

- Used pre-trained fasttext embedding, these were feed into a Stacked Bi-directional GRU layer.

- They both were combined and predicted the next word till the end of caption using greedy search (during testing).

## Result

#### Weights

## To Do 

- Add attention

- Use beam search instead of greddy seaech