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..
Lattice Boltzmann Model for Learning Real-World Pixel Dynamicity
Authors: Guangze Zheng, Shijie Lin, Haobo Zuo, Si Si, Ming-Shan Wang, Changhong Fu, Jia Pan
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comprehensive evaluations of real-world point tracking benchmarks such as TAP-Vid and Robo TAP validate LBM s efficiency. A general evaluation of large-scale open-world object tracking benchmarks such as TAO, BFT, and OVT-B further demonstrates LBM s real-world practicality.4 Experiments |
| Researcher Affiliation | Collaboration | 1HKU 2Institute of Zoology, CAS 3Tongji University 4Lim X Dynamics |
| Pseudocode | No | The paper describes the model architecture and processes using figures, equations, and descriptive text, but does not include any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code for training and evaluation is provided. The links to public datasets are also provided. |
| Open Datasets | Yes | TAP-Vid Kubric Greff et al. (2022) dataset for training, which contains 11k video sequences of 24 frames each. The training process encompasses 150 epochs (approximately 100k iterations) using 4 NVIDIA H800 GPUs with a total batch size of 16 and FP16 mixed-precision.Evaluation datasets Three real-world point tracking benchmarks are employed, including TAP-Vid DAVIS, TAP-Vid Kinetics, and Robo TAP Vecerik et al. (2024). Open-world object tracking datasets include: TAO Dave et al. (2020) validation set, BFT Zheng et al. (2024) test set, and OVT-B Liang and Han (2024). |
| Dataset Splits | Yes | TAP-Vid Kubric Greff et al. (2022) dataset for training, which contains 11k video sequences of 24 frames each.The training process encompasses 150 epochs (approximately 100k iterations) using 4 NVIDIA H800 GPUs with a total batch size of 16 and FP16 mixed-precision.Evaluation datasets Three real-world point tracking benchmarks are employed, including TAP-Vid DAVIS, TAP-Vid Kinetics, and Robo TAP Vecerik et al. (2024). Open-world object tracking datasets include: TAO Dave et al. (2020) validation set, containing 988 videos spanning 330 object categories annotated at 1 frame per second; BFT Zheng et al. (2024) test set, comprising 36 videos featuring highly dynamic avian objects; and OVT-B Liang and Han (2024), a large-scale open-world object tracking benchmark encompassing 1,973 videos with 1,048 object categories. |
| Hardware Specification | Yes | LBM demonstrates real-time operational capability at 14.3 FPS on the NVIDIA Jetson Orin NX Super edge platform, with a 3.9 speed advantage over Track-On, thereby showing computational efficiency and practicality in real-world environments. The training process encompasses 150 epochs (approximately 100k iterations) using 4 NVIDIA H800 GPUs with a total batch size of 16 and FP16 mixed-precision. The speeds are tested on an NVIDIA Jetson Orin NX super. |
| Software Dependencies | No | The paper mentions software components like "Adam W optimizer", "FP16 mixed-precision", "ONNX format", and "Tensor RT quantization", but does not provide specific version numbers for these or other key software dependencies. |
| Experiment Setup | Yes | We employ the Adam W optimizer with a peak learning rate of 5 10 4 and weight decay of 1 10 5, implementing a cosine decay schedule with 5% linear warm-up initialization. The whole training process takes over 2 days on 4 NVIDIA H800 GPUs with 4 batches each. LBM adopts identical data augmentation strategies as Co Tracker Karaev et al. (2024b), processing input images at 384 512 resolution while sampling 256 points per batch. |