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..
TrackingWorld: World-centric Monocular 3D Tracking of Almost All Pixels
Authors: Jiahao Lu, Weitao Xiong, Jiacheng Deng, Peng Li, Tianyu Huang, Zhiyang Dou, Cheng Lin, Sai-Kit Yeung, Yuan Liu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive evaluations on both synthetic and real-world datasets demonstrate that our system achieves accurate and dense 3D tracking in a world-centric coordinate frame. ... 4 Experiment ... 4.2 Quantitative comparisons ... 4.4 Ablation study |
| Researcher Affiliation | Academia | 1HKUST 2USTC 3CUHK 4HKU 5XMU 6MUST |
| Pseudocode | No | The paper describes methods using natural language and mathematical equations, but does not present any explicitly labeled pseudocode blocks or algorithms. |
| Open Source Code | Yes | https://github.com/IGL-HKUST/TrackingWorld |
| Open Datasets | Yes | We evaluate camera pose estimation performance on three dynamic datasets: Sintel [46], Bonn [47], and TUM-D [48]. ... To evaluate the performance of 3D sparse tracks, we conduct experiments on two datasets, ADT [57] with moving cameras, and PStudio [58] with static cameras. ... We evaluate the dense 2D tracking performance on the CVO [60] test set... Fig. 3 qualitatively visualizes the world-centric dense tracking results produced by our method on the DAVIS [61] dataset. |
| Dataset Splits | No | For ADT, we sample one video every 100 videos, and for PStudio, one video every 20 videos. ... For all three datasets, we adopt the same settings as Mon ST3R [37]. ... Each subset contains approximately 500 videos with 7 frames. The paper describes sampling strategies and refers to settings from other papers, but does not provide explicit training/test/validation splits for its own experimental setup. |
| Hardware Specification | Yes | All experiments are conducted on an RTX 4090 GPU. |
| Software Dependencies | No | We use Co Tracker V3 [1] and DELTA [4] to obtain dense tracking results, and adopt Uni Depth [23] as the depth prior. For the dynamic mask estimation method, we follow Uni4D [24] to apply VLM [41] and Grounding SAM [42, 43] to segment out foreground dynamic objects. Alternatively, we could also choose the Segment Any Motion [44] to get dynamic masks. The paper mentions software tools and models but does not provide specific version numbers for them or other key software components like programming languages or libraries. |
| Experiment Setup | No | More details about hyperparameters can be found in the supplementary materials. The paper states that hyperparameters are available in supplementary materials, but does not provide specific values or training configurations in the main text. |