Comparing Evaluation Protocols on the KTH Dataset

67 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Lifelong learning of human actions with deep neural network self-organization (2017, methodology): “Our result is competitive with the two best results: 95.6% by Ravanbakhsh et al. (2015) and 95.04% by Gao et al. (2010).”
Fast Exact Hyper-graph Matching with Dynamic Programming for Spatio-temporal Data (2014, unknown): “We would like to point out that although many research results have been published on the KTH database, most of these results cannot be directly compared due to their differing evaluation protocols, as has been indicated in the detailed report on the KTH database [18].”
Robust human action recognition via long short-term memory (2013, methodology): “of the KTH dataset into the training and test sets, we cannot conclusively state that our improvement is due to differences in the network architecture or training regime, as opposed to partition choice (where the latter has been shown to affect performance [8]).”
Sequential Deep Learning for Human Action Recognition (2011, methodology): “We also notice that the first step of our model gives relatively stable results on the 5 configurations, compared to the fluctuations generally observed for the other methods [4].”