GSM: Graph Similarity Model for Multi-Object Tracking

Authors: Qiankun Liu, Qi Chu, Bin Liu, Nenghai Yu

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments on challenging MOT benchmarks and the experimental results demonstrate the effectiveness of the proposed method.
Researcher Affiliation Academia Qiankun Liu , Qi Chu , Bin Liu and Nenghai Yu University of Science and Technology of China, China liuqk3@mail.ustc.edu.cn, {qchu, flowice, ynh}@ustc.edu.cn
Pseudocode No The paper does not include pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not contain any explicit statement about releasing source code or a link to a code repository.
Open Datasets Yes Experiments are conducted on MOT Benchmarks, including MOT16 and MOT17 [Milan et al., 2016].
Dataset Splits Yes The 7 sequences in MOT16 train split are divided into train set and validation set to conduct ablation study. Validation set: MOT16-09 and MOT16-10. Training set: the rest sequences in MOT16 train split.
Hardware Specification Yes Evaluation is on a workstation with 2.6 GHz CPU and Nvidia TITAN Xp GPU.
Software Dependencies No The paper states:
Experiment Setup Yes All image patches are resized to 64x128. The appearance CNN, RPE and binary classifer are trained end-to-end with binary cross entropy loss for 30 epochs. The learning rate is initialized as 0.002 and decades every 10 epochs with exponential decay rate 0.5. Online hard example mining (OHEM) was adopted to address the imbalance of positive/negative issue. ... 64 candidate bounding-boxes are sampled for each lost object.