Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
EAG3R: Event-Augmented 3D Geometry Estimation for Dynamic and Extreme-Lighting Scenes
Authors: Xiaoshan Wu, Yifei Yu, Xiaoyang Lyu, Yihua Huang, Bo Wang, Baoheng Zhang, Zhongrui Wang, Xiaojuan Qi
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate that EAG3R significantly outperforms state-of-the-art RGB-only baselines across monocular depth estimation, camera pose tracking, and dynamic reconstruction tasks. |
| Researcher Affiliation | Academia | 1The University of Hong Kong 2Southern University of Science and Technology |
| Pseudocode | No | The paper describes methods through equations and textual explanations, but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | The paper will provide open access to the data and code necessary to reproduce the main experimental results. It also includes sufficient instructions in the supplemental material on how to faithfully replicate the experiments conducted in the paper. |
| Open Datasets | Yes | We evaluate EAG3R on the MVSEC dataset [78]... Given the scarcity of such data, we selected the Multi Vehicle Stereo Event Camera (MVSEC) dataset [78]. It provides synchronized stereo events and reliable Li DAR-derived depth GT. |
| Dataset Splits | Yes | All models were trained exclusively on MVSEC s outdoor_day2 sequence (normal daylight) and tested on the challenging outdoor_night1-3 sequences (extreme low-light). |
| Hardware Specification | Yes | The training process completes in approximately 24 hours on 4 NVIDIA RTX 3090 GPUs. ...as confirmed by our experiments running on an NVIDIA A100 GPU. |
| Software Dependencies | No | The paper mentions optimizers (AdamW, Adam) but does not provide specific version numbers for any key software components or libraries (e.g., Python, PyTorch, TensorFlow, CUDA versions). |
| Experiment Setup | Yes | Fine-tuning is performed for 25 epochs, using 8,000 image-event pairs per epoch. We employ the Adam W optimizer with a learning rate of 5 10 5 and a mini-batch size of 4 per GPU. ... For global optimization, we adopt the same setting as Mon ST3R, with hyperparameters wsmooth = 0.01, wflow = 0.01, and wevent_base = 0.01. We use the Adam optimizer for 300 iterations with a learning rate of 0.01. |