Homography Decomposition Networks for Planar Object Tracking

Authors: Xinrui Zhan, Yueran Liu, Jianke Zhu, Yang Li3234-3242

AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments show that our proposed approach outperforms the state-of-the-art planar tracking methods at a large margin on the challenging POT, UCSB and POIC datasets. Experiment Setup We conducted experiments on a PC with an intel E5-2678-v3 processor (2.5GHz), 32GB RAM and Nvidia GTX 2080Ti GPU.
Researcher Affiliation Collaboration 1 Zhejiang University 2 Alibaba-Zhejiang University Joint Institute of Frontier Technologies 3 East China Normal University
Pseudocode No No pseudocode or algorithm blocks found in the paper.
Open Source Code Yes Codes and models are available at https://github.com/zhanxinrui/HDN.
Open Datasets Yes The algorithm randomly chooses x from a certain range to perform the transformation on COCO14 (Lin et al. 2014). To solve the unrealistic problem of synthetic datasets for residual component training, we sample the images of tracking dataset GOT10k (Huang, Zhao, and Huang 2019) with a small interval threshold as training data and adopt an unsupervised residual loss L Λ+.
Dataset Splits No No explicit training/validation/test split percentages or sample counts for validation are provided. The paper mentions training on COCO14 and GOT10k, and testing on POT, UCSB, and POIC, but doesn't detail specific validation splits.
Hardware Specification Yes We conducted experiments on a PC with an intel E5-2678-v3 processor (2.5GHz), 32GB RAM and Nvidia GTX 2080Ti GPU.
Software Dependencies No The paper states, 'Our proposed method is implemented in Pytorch,' but does not provide a specific version number for PyTorch or any other software dependencies.
Experiment Setup Yes For the hyperparameters of HDN in training and testing, we set λ1 = 100 , λ2 = 1, λ3 = 1, λ4 = 0.25, K = 100. γ [1/1.38, 1.38], θ [ 0.7, 0.7], t [ 32, 32], k1 [ 0.1, 0.1], k2 [ 0.015, 0.015], ν [ 0.0015, 0.0015]. The probability threshold τ in Lcls is set to 0.7.