Automated Lip reading using convolutional Neural Networks in python

Saqhas


1

Vote

Description:

Problem description:

Deep learning algorithms have shown great results in speech recognition domain, So here we have used deep learning techniques to enable the machines to read the lips from a video without sound better than humans. By analysing the movement of lips of a person we are trying to predict what that person is trying to speak. Automated Lip reading can be helpful in many ways. Some of them are:

  • Silent dictation in public spaces.
  • Covert conversation.
  • Helping the people with speaking ade in talking to other people.
  • Improved hearing aids.
  • Speech recognition in a noisy environment.

The talk will be focused on :

  • How the problem should be tackled.
  • Discussion of different phases
  • Algorithms and python libraries used for implementation.

Prerequisites:

The session is fully for beginner's who have just entered the field of machine learning and deep learning. Every small detail will be covered in the talk. Beginner's knowledge of the following items would be helpful.

  • Knowledge of Numpy: A good resource for the same is : Udemy Deep learning Stack
  • Basic understanding of OpenCV: A good resource for the same is:Udemy OpenCV Basics. This much is enough, we would also be covering the important content in the talk.
  • Basic Knowledge of Convolutional Neural Networks: An excellent resource to understand this is CNN by Datacamp. The motive should be to understand the basic working of Convolutional Neural Networks. The coding part will be explained in the talk as well relating each step to it's working.
  • Knowledge of basic metrics: These are ways to evaluate the performance of our model. An excellent resource is:Metrics for Evaluations
  • Basic understanding of Recurrent Neural Networks: An excellent resource to understand this is Understanding LSTM Networks. Similar to CNN the motive should be to understand the basic working of Recurrent Neural Networks. The coding part will be discussed in the talk.

Content URLs:

The GitHub repository and the talk slide are:

  • Slides: Will be updated soon.
  • Github repo: Will be updated soon.

Speaker Info:

The talk will be presented by 3 speakers who are presently 4th-year students from Jaypee Institute of information technology, Noida. They are machine learning enthusiasts and researchers. They had been working in this field for the past one-half year and have worked on a variety of projects ranging from customer segmentation, recommendations to the projects from the field of computer vision, deep learning and reinforcement learning. They have also worked with an AI startup where they had to predict customer churn.

Speaker Links:

The LinkedIn Profile are:

The Github Profile are:

Section: Data science
Type: Talks
Target Audience: Intermediate
Last Updated: