Rekordio - LWT and Presenc IO

Name of the project: LWT and Presenc IO

Company introduction: Rekordio is a US/Australian-based organization that ensures every spoken word makes its mark and contributes to something great. The company uses AI/ML to distill your conversations into meaningful data and put your words to work for you. They are working on developing a new stealth communication platform that is coming soon to the global markets.

Non-Technical View of the project: We had to build a web application or video conferencing tool like zoom. We worked on making a video calling web application that would utilize low bandwidth that is for internal communication via video calls between the employees of the company resulting in clarity for the audio as well as video conferencing. It has the functionality to arrange meetings over another platform with its built support in a calendar for meeting and agenda managers.UI side is used to display insights information gathered by AI model from meeting videos and also manage meetings and agendas.

Technical View of the Project: It consists of the Frontend and backend using Machine learning and Deep Learning. The concept was to send just the face motion data with that user's image to the other end where the user on that end would be able to see the motions and the image integrated in such a way that it won't seem like it isn't a video stream. But technically, they don't send the video stream to the other end. Hence it reduces the usage of internet/bandwidth by reducing the amount of data that is needed to be transferred. The task was to make this model's architecture small and maintain the accuracy similar to the original model and also increase the fps rate of the model.

This was the core task in the backend and the other backend work was to use the machine/deep learning end where the developers tried to use a deep learning model to encode videos into a latent vector compression and then send that encoded version to the other end, this way we could compromise on the quality a bit save bandwidth at the same time On the front end, the task was to design the API for Private Call where the user can call anyone who is there contact list/Group and Build the backend for the same. A call can be placed if the person is not busy or not on another call. Send the notification to the sender if the recipient is busy. While the other front-end task was to make the user interface work with the webcam and fetch the video stream frames and send it to the backend for our model to process this data and produce the output. This output will be received by the user on the other end.

Category of the project: Machine Learning with Deep Learning

Duration of the Project: Around 4 months

Form of Project Created: A web application like zoom or any other was built by the team which consists of Front-End as well as Back-End