Overview / Usage

PhoBot is image captioning web app. This app can be used to upload an image and get the relevant captions for it.

Methodology / Approach

I gathered dataset from Flickr30k dataset which has atleast 5 captions for each image. It has total 30k images of different types. I used InceptionV3 model which is pertained on imagenet. I removed last 2 layers and passed it to a final model. I used GloVe Vectoriser to train text i.e captions.

I made final model to take inputs from both, text and images. The output were sequence prediction from the final model.

Technologies Used

Keras

Tensorflow

InceptionV3 Pretrained Model

Glove Vector

Repository

https://github.com/AshiishKarhade/Photobot-Image-Captioning