Skip to content

Uses a CNN Encoder and a RNN Decoder to generate captions for input images.

Notifications You must be signed in to change notification settings

eljandoubi/Image-Captioning

Repository files navigation

Image-Captioning

Uses a CNN Encoder and a RNN Decoder to generate captions for input images.

Here's a sumary of the steps involved.

  • Dataset used is the COCO data set by Microsoft.
  • Feature vectors for images are generated using a CNN based on the ResNet architecture by Google.
  • Word embeddings are generated from captions for training images. NLTK was used for working with processing of captions.
  • Implemented an RNN decoder using LSTM cells.
  • Trained the network for nearly 9 hrs using GPU to achieve average loss of about 2.8.
  • Obtained outputs for some test images to understand efficiency of the trained network.

Alt

About

Uses a CNN Encoder and a RNN Decoder to generate captions for input images.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published