3D-reconstruction-using-NeRF-from-2D-images-or-video

The aim of my project was to develop a user-friendly method for generating realistic 3D models and displaying them on smartphones using augmented reality. The only input required from the user is a short video of the desired object, which can be uploaded to the app. The project uses a modified version of Nvidia's Instant-NeRF model, along with some pre-processing steps, to improve the quality and accuracy of the reconstructed 3D models and complete it within 5 to 10 minutes.

Demo

https://drive.google.com/file/d/1bEckQom2CcpFmc6btTXutiNDQtV8oxjC/view?usp=sharing

Installation

Need libraries

• CUDA

• Python version 3.1 or higher

• CMake build of Instant NGP (with necessary prerequistes)

• COLMAP

• YOLOV8

• U2-Net

• Open3D

• OpenCv

• pyTorch

• flutter

• flask

• AR core

Screenshots

Project Flow Results Samples Results Samples Results Samples

Reasearch paper

Reasearch Paper

Acknowledgements

I would like to thank Dr. Mohammed A.-Megeed Salem, my thesis supervisor, for his valuable knowledge and guidance throughout this project. I would also like to acknowledge ENG. Ramez Ibrahim, my mentor, for his helpful advice and feedback that helped me achieve my goals and improve my skills.

The project is based of the Nvidia's Instant-NeRF model:

• Nvidia's instant NGP