Sampling-Resilient Multi-Object Tracking

Authors: Zepeng Li, Dongxiang Zhang, Sai Wu, Mingli Song, Gang Chen

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on three benchmark datasets show that our proposed tracker achieves the best trade-off between efficiency and accuracy.
Researcher Affiliation Academia 1 The State Key Laboratory of Blockchain and Data Security, Zhejiang University 2 College of Computer Science and Technology, Zhejiang University
Pseudocode No The paper describes the proposed methods using textual explanations and mathematical equations, but it does not include any pseudocode blocks, algorithms, or flowcharts labeled as such.
Open Source Code No The paper mentions using YOLOX provided by previous trackers and comparing against 'open-sourced trackers,' but it does not provide an explicit statement about releasing its own source code for the proposed SR-Track methodology or a link to such a repository.
Open Datasets Yes We use three benchmark datasets for performance evaluation, including MOT17 (Milan et al. 2016), MOT20 (Dendorfer et al. 2020) and Dance Track (Sun et al. 2022).
Dataset Splits Yes Dance Track is a recent dataset proposed to emphasize the importance of motion analysis. ... It provides 100 videos and the split ratio for training, validation and test dataset is 40 : 25 : 35.
Hardware Specification Yes All the experiments are conducted using Py Torch and ran on a desktop with 10th Intel(R) Core(TM) i9-10980XE CPU @ 3.00GHz and NVIDIA Ge Force RTX 3090Ti GPU.
Software Dependencies No The paper mentions 'Py Torch' as the framework used for experiments but does not specify a version number for PyTorch or any other software dependencies.
Experiment Setup Yes As to our proposed Kalman filter, we set hidden size to 128 for the LSTM network and adopt two-layer Bayesian neural network to implement Q and R. All models are trained using the Adam optimizer for 100 epochs with a batch size of 32. The initial learning rate is set to 0.01 and linearly decayed to 0.0001.