International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 09 Issue: 05 | May 2022
p-ISSN: 2395-0072
www.irjet.net
Real Time Head Generation for Video Conferencing Prof. Sumedha Ayachit1, Rohan Sabale2, Avinash Parit3, Shreya Shere4 , Varun Raikar5 1Prof.
Dept. of Information Technology, JSCOE, Pune, Maharashtra, India of Information Technology, JSCOE, Pune, Maharashtra, India Department of Information Technology, Jayawantrao Sawant College of Engineering, Pune ---------------------------------------------------------------------***--------------------------------------------------------------------2,3,4,5 Dept.
Abstract - Video conferences area unit receiving genuine
The driving force of our new modern era has become the new and advanced technology. Hence, using technology like WebRTC and FOMM, we developed an advanced Video Conferencing System which can even be used in lower bandwidth and poor network connections which will help people facing these issues. This paper mainly focuses on outlining and executing a web-based conferencing system by initiating peer to peer connection using Web-RTC and face reconstruction and reformation by the facial data point already extracted using First Order Motion Model. This video conferencing system will help connect people who are having poor digital Infrastructure and cannot afford high speed data connection.
interest as a technique of interaction as web conferencing has improved dramatically in the last few years. WebRTC technology uses a high-speed data transmission channel to establish communication between users. We propose a realtime talking-head video synthesis model for video conferencing. Our model reconstructs a source video at the receiver side to maintain a steady and lag-free experience for face-to-face video conferencing. It uses less bandwidth compared to the commercial H.264 standard. It extracts and retargets motion from the sender video frame by frame and synthesis video on the receiver end using the first image and motion keypoints. It successfully takes out facial expressions, head poses, and eye movements from the faces. It synthesizes the talking-head from the original viewpoint of the source image. Key Words: WebRTC
1.1 Background and Related Work The overall design of the WebRTC system is arranged out in [1] [2], and therefore the data is pictured to flow in a very peer-to-peer fashion directly between the two net browsers. Therefore the protocols used to transport, communicate and secure the encrypted media are specified. The main points relating to the WebRTC information channels square measure arranged clearly. The problems associated with the transport layers unit of measurement revolve around NAT (Network Address Traversal) and firewall traversal, multiplexing of information and media over one transport flow is additionally addressed.
Video conferencing, Machine Learning,
1. INTRODUCTION Digital property composed of wireless, satellite technologies, and wired technologies these are the utility of the twenty first century. Digital life helps us improve many acreages of our lives like virtual education, monitoring health, and keeping physicians informed using internet technology and other aspects of life.
Earlier works on deep video generation are mentioned. However, Spatio-temporal neural networks get render video frames from noise vectors [6]. A lot of recently, many approaches tackled the matter of conditional video generation. For example, Wang et al. [7] combine a recurrent neural network with a VAE to get face videos. Considering a more comprehensive variety of applications, Tulyakov et al. [8] introduced MoCoGAN, a continual design adversarially trained to synthesize videos from noise, categorical labels, or static pictures.
Due to COVID-19 pandemic there is an unprecedented need of cyber technologies and its related online services. One of the main key eventful changes concerned was how individuals and communities socialize and communicate with each another. Virtual proceedings have become the most sort of top of the list way of communicating and holding meeting or teaching all over the world from businesses to schools and colleges as they can communicate from the comfort of homes. Like current one of the biggest video conferencing service providers, Zoom had usage of 10 million daily meeting participant’s pre covid which are far lesser than now. Post Covid they have now hundreds of million daily participants. But for many people, the increased usage of this growing tech has been troublesome. It has many challenges, like digital infrastructure, affordability of data connection, and quality internet access.
© 2022, IRJET
|
Impact Factor value: 7.529
X2Face [5] uses a dense motion field which is extracted by the image to get the output video via image distortion. Equally to us, they use a reference create that's wont to get a canonical illustration of the article. In our formulation, we don't need an exact connection to make, resulting in considerably less complicated optimization and improved image quality.
|
ISO 9001:2008 Certified Journal
|
Page 335