# GARF: Gaussian Activated Radiance Fields for High Fidelity Reconstruction and Pose Estimation

**Authors:** Shin-Fang Chng, Sameera Ramasinghe, Jamie Sherrah, Simon Lucey  
**arXiv:** [2204.05735](https://arxiv.org/abs/2204.05735)  
**Date:** 2022-04-12  
**Summary page:** https://www.summarizepaper.com/en/paper/2204.05735v1/

## Key points

- Photorealistic novel views synthesis of real-world scenes has shown remarkable results
- Existing approaches often require accurate prior camera poses, posing a significant limitation
- GARF is an innovative architecture that eliminates the need for positional embeddings
- GARF leverages Gaussian activations to enhance reconstruction quality and pose estimation accuracy
- Research by Shin-Fang Chng, Sameera Ramasinghe, Jamie Sherrah, and Simon Lucey demonstrates GARF's superiority over current techniques
- GARF sets a new standard for achieving exceptional results in complex scene synthesis tasks

## Summary

In the realm of photorealistic novel views synthesis of real-world scenes,  have demonstrated remarkable results. However, a significant limitation of most existing approaches is the necessity for accurate prior camera poses. While some methods like BARF exist for jointly recovering the radiance field and camera pose, they often rely on a cumbersome coarse-to-fine auxiliary positional embedding to ensure optimal performance. In light of these challenges, a groundbreaking solution has emerged in the form of . This innovative architecture eliminates the need for positional embeddings, instead leveraging Gaussian activations to enhance reconstruction quality and pose estimation accuracy. The research conducted by Shin-Fang Chng, Sameera Ramasinghe, Jamie Sherrah, and Simon Lucey showcases GARF's superiority over current state-of-the-art techniques in terms of high fidelity reconstruction and precise pose estimation. By introducing GARF as a positional embedding-free alternative, this study not only pushes the boundaries of neural radiance field technology but also sets a new standard for achieving exceptional results in complex scene synthesis tasks. The project page provided offers further insights into the development and implementation of GARF, highlighting its potential impact on advancing the field of computer vision and graphics.