Articulated Pose Estimation Using Hierarchical Exemplar-Based Models

Authors: Jiongxin Liu, Yinxiao Li, Peter Allen, Peter Belhumeur

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our method extensively on multiple benchmarks, and conduct diagnostic experiments to show the effect of different components in our method.
Researcher Affiliation Academia Jiongxin Liu, Yinxiao Li, Peter Allen, Peter Belhumeur Columbia University in the City of New York {liujx09, yli, allen, belhumeur}@cs.columbia.edu
Pseudocode Yes Algorithm 1: Inference Procedure for Pose Estimation
Open Source Code No The paper does not provide any statement about releasing source code or a link to a code repository.
Open Datasets Yes LSP dataset (Johnson and Everingham 2010) includes 1, 000 images for training and 1, 000 images for testing... CUB-200-2011 bird dataset, which contains 5, 994 images for training and 5, 794 images for testing.
Dataset Splits Yes To avoid over-fitting, the training is conducted on a held-out validation set that was not used to train the DCNNs.
Hardware Specification No The paper does not explicitly describe the hardware (e.g., specific GPU or CPU models) used for running the experiments.
Software Dependencies No The paper mentions using "Caffe (Jia et al. 2014)" but does not provide specific version numbers for Caffe or any other software libraries or dependencies.
Experiment Setup Yes We augment the training data by left-right flipping, and rotating through 360 ... In this experiment, we build a hierarchy of four levels for human body... As the bird body is relatively more rigid than the human body, the degrees of bird s internal nodes can be larger, resulting in fewer levels... we design a three-level hierarchy for birds... To gain an understanding of the effect of the components of our inference algorithm, we evaluate our full method (which will be referred to as Ours-full ), and variants of our method (which will be referred to as Ours-partial , and Ours-no-HIER ).