Action Recognition From Skeleton Data via Analogical Generalization Over Qualitative Representations

Authors: Kezhen Chen, Kenneth Forbus

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Results on three public datasets illustrate its utility.
Researcher Affiliation Academia Kezhen Chen, Kenneth D. Forbus Northwestern University Kezhen Chen2021@u.northwestern.edu
Pseudocode No The paper describes the pipeline and steps in prose and with diagrams, but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statement or link indicating that the source code for the described methodology is open-source or publicly available.
Open Datasets Yes We evaluated dynamic feature selection on the Florence 3D Action dataset (Seidenari et al. 2013) and The University of Texas at Dallas Multimodal Human Action Dataset (UTD-MHAD) was collected as part of research on human action recognition by Chen et al. (Chen et al. 2015). To further evaluate our method, we ran an experiment on the UTKinect-Action3D dataset (Xia, Chen, and Aggarwal 2012).
Dataset Splits Yes Cross-subject validation was used. We follow the leave one out cross validation protocol (LOOCV) (Seidenari et al. 2013) to better compare with other methods.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions software components like Cog Sketch and QSRlib, but does not provide specific version numbers for any of its software dependencies.
Experiment Setup Yes For all experiments reported here, we used an assimilation threshold of 0.7. We used a probability cutoff of 0.2 in all experiments. The search stops either when a cutoff is reached (here, the cutoff is four optional features, which provides a reasonable tradeoff between accuracy and efficiency)