/image-description

This repo represents our machine learning project Image Description which is used to generate a description of an image based on activities and objects detected in the image.

Primary LanguageCSSMIT LicenseMIT

Image Description

ID logo

Image Description refers to the process of generating meaningful descriptions from images based on objects and activities detected in the images.

Specifications

  • Dataset : MS COCO 2014

  • Dataset link : https://cocodataset.org/

  • Image Datasize : 1,25,000

  • Trainable Parameters : 13,855,013

  • Layers : 12


Model Architecture :

Model


To run this project on local machine

  • Install virtualenv by:

pip install virtualenv

  • Create virtual environment :

virtualenv my_env

  • Activate virtual env :

my_env\Scripts\activate

  • Install requirements for project :

pip install -r requirements.txt

  • Run this project :

python app.py

  • Test the project :

Go to localhost:9999 in your browser.


Live Version

You can see it on https://img-descriptor.herokuapp.com/


Output

Output 1

Output 2


Creators