# Do We Really Need Scene-specific Pose Encoders?

arXiv: https://arxiv.org/abs/2012.12014 · HTML: https://mlsanity.com/paper/2012.12014/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 22 Dec 2020.

## Built on by (most cited first, 10 of 10)

- [Camera Pose Auto-Encoders for Improving Pose Regression](https://mlsanity.com/paper/2207.05530/) (2022)
- [RobustLoc: Robust Camera Pose Regression in Challenging Driving Environments](https://mlsanity.com/paper/2211.11238/) (2022)
- [Coarse-to-Fine Multi-Scene Pose Regression with Transformers](https://mlsanity.com/paper/2308.11783/) (2023)
- [Benchmarking Visual-Inertial Deep Multimodal Fusion for Relative Pose Regression and Odometry-aided Absolute Pose Regression](https://mlsanity.com/paper/2208.00919/) (2022)
- [Activating Self-Attention for Multi-Scene Absolute Pose Regression](https://mlsanity.com/paper/2411.01443/) (2024)
- [A Lightweight Domain Adaptive Absolute Pose Regressor Using Barlow Twins Objective](https://mlsanity.com/paper/2211.10963/) (2022)
- [HyperPose: Hypernetwork-Infused Camera Pose Localization and an Extended Cambridge Landmarks Dataset](https://mlsanity.com/paper/2303.02610/) (2023)
- [NeuroLoc: Encoding Navigation Cells for 6-DOF Camera Localization](https://mlsanity.com/paper/2505.01113/) (2025)
- [Robust Visual Localization via Semantic-Guided Multi-Scale Transformer](https://mlsanity.com/paper/2506.08526/) (2025)
- [Relative Pose Regression with Pose Auto-Encoders: Enhancing Accuracy and Data Efficiency for Retail Applications](https://mlsanity.com/paper/2508.10933/) (2025)

## Abstract (arXiv)

Visual pose regression models estimate the camera pose from a query image with a single forward pass. Current models learn pose encoding from an image using deep convolutional networks which are trained per scene. The resulting encoding is typically passed to a multi-layer perceptron in order to regress the pose. In this work, we propose that scene-specific pose encoders are not required for pose regression and that encodings trained for visual similarity can be used instead. In order to test our hypothesis, we take a shallow architecture of several fully connected layers and train it with pre-computed encodings from a generic image retrieval model. We find that these encodings are not only sufficient to regress the camera pose, but that, when provided to a branching fully connected architecture, a trained model can achieve competitive results and even surpass current textitstate-of-the-art pose regressors in some cases. Moreover, we show that for outdoor localization, the proposed architecture is the only pose regressor, to date, consistently localizing in under 2 meters and 5 degrees.