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..

RGB-Only Supervised Camera Parameter Optimization in Dynamic Scenes

Authors: Fang Li, Hao Helen Zhang, Narendra Ahuja

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We perform experiments on 4 real-world datasets (Ne RF-DS, DAVIS, i Phone, and TUM-dynamics) and 1 synthetic dataset (MPI-Sintel), demonstrating that our method estimates camera parameters more efficiently and accurately with a single RGB video as the only supervision. We present exhaustive quantitative and qualitative experiments and extensive ablation studies that demonstrate the superior performance of our proposed method and the contribution of each component.
Researcher Affiliation Academia Fang Li University of Illinois at Urbana-Champaign Champaign, IL 61820 EMAIL Hao Zhang University of Illinois at Urbana-Champaign Champaign, IL 61820 EMAIL Narendra Ahuja University of Illinois at Urbana-Champaign Champaign, IL 61820 EMAIL
Pseudocode No The paper describes the methods in text (Section 3) and mathematical formulas but does not include any clearly labeled 'Pseudocode' or 'Algorithm' block.
Open Source Code No Answer: [No] Justification: We will release the code upon the acceptance of the paper.
Open Datasets Yes We evaluate the performance of our method through extensive experiments on 5 popular public datasets Ne RF-DS [50], DAVIS [28], i Phone [7], MPI-Sintel [2], and TUM-dynamics [36], demonstrating our superior performance.
Dataset Splits Yes Regarding the train/test split of the NVS evaluation, for every 2 adjacent frames, we take the first frame for training and the second frame for testing.
Hardware Specification Yes The optimization is conducted on 1 NVIDIA A100 40GB GPU
Software Dependencies No The paper mentions 'Adam [27] optimizer' and that it builds 'our patch-wise tracking filters on Co Tracker [13]', but specific version numbers for these or other software libraries are not provided.
Experiment Setup Yes The optimization is conducted on 1 NVIDIA A100 40GB GPU with Adam [27] optimizer and learning rates l Q = 0.01, lt = 0.01, lf = 1.0, l Pcali = 0.01, and lฮ“raw = 0.01. We also choose to build our patch-wise tracking filters on Co Tracker [13] and load its pre-training weights. The hyperparameters of our patch-wise tracking filters are set at ฯ„var = 0.1, B = 100, w Ne RF-DS, DAVIS, MPI-Sintel = 12, and wi Phone, TUM = 24. Notably, w is only related to the frame size. Besides, throughout our experiments, we have 200 and 50 iterations in Stage1 and Stage2 respectively.