Chirality Nets for Human Pose Regression

Authors: Raymond Yeh, Yuan-Ting Hu, Alexander Schwing

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate chirality nets on the task of human pose regression... We demonstrate the generalization and effectiveness of our approach on three pose regression tasks over four datasets... Our approach achieves/matches state-of-the-art results...
Researcher Affiliation Academia Raymond A. Yeh , Yuan-Ting Hu*, Alexander G. Schwing Department of Electrical Engineering, University of Illinois at Urbana-Champaign {yeh17, ythu2, aschwing}@illinois.edu
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes The Pytorch implementation and unit-tests of the proposed layers are part of the supplementary material. We have also included a short Jupyter notebook demo to illustrate the key concepts.
Open Datasets Yes We evaluate on two standard datasets, the Human3.6M [22] and the Human Eva I [49].
Dataset Splits Yes We use the same train and test subject splits.
Hardware Specification No The paper states, 'We thank NVIDIA for providing GPUs used for this work and Cisco for access to the Arcetri cluster,' but it does not specify any particular GPU models, CPU models, memory details, or detailed cluster specifications.
Software Dependencies No The paper mentions 'Pytorch implementation' and uses optimizers like 'Adam' and 'SGD,' and tools like 'Open Pose,' but it does not provide specific version numbers for these software components or libraries.
Experiment Setup Yes Our model follows the supervised training procedure and network design of Pavllo et al. [42]. Our network is the identical temporal convolutional network architecture, where each layer is replaced with its chiral version, i.e., 1D dilated convolution, batch-normalization, and dropout layers. We also replace ReLU non-linearities with Tanh to achieve equivariance.