Learning Precise Temporal Point Event Detection with Misaligned Labels

Authors: Julien Schroeter, Kirill Sidorov, David Marshall9505-9514

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 6 Experiments In order to demonstrate the effectiveness and flexibility of our approach, a broad range of challenging experiments are conducted.
Researcher Affiliation Academia Julien Schroeter, Kirill Sidorov, David Marshall School of Computer Science & Informatics, Cardiff University, UK {Schroeter J1, Sidorov K, Marshall AD}@cardiff.ac.uk
Pseudocode No The paper does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Code is available1. 1 https://github.com/Schroeter Julien/AAAI-2021-Learning Precise-Temporal-Point-Event-Detection-with-Misaligned-Labels
Open Datasets Yes Golf DB: A Video Database for Golf Swing Sequencing (Mc Nally et al. 2019), puff Marker dataset (Saleheen et al. 2015), MAPS database (Emiya, Badeau, and David 2010), IDMT-SMT-Drums dataset (Dittmar and Gärtner 2014)
Dataset Splits Yes The (4-fold) cross-validated mean accuracy is reported. (Table 1 caption) and The results, produced using ten 6-fold (leave-one-patient-out) cross-validations are summarized in Table 2.
Hardware Specification No We gratefully acknowledge the support of NVIDIA Corporation with the donation of a GPU used for this research. (This only specifies 'a GPU' from NVIDIA, not a specific model, memory, or other hardware details.)
Software Dependencies No The paper mentions the 'mir eval library' (Raffel et al. 2014) but does not provide specific version numbers for it or any other software dependencies.
Experiment Setup Yes (SM = 100ms, ατ = max(min( τ 105 , .9), .2).) (Section 6.3) and (learning rate: 10 4, batch size: 32, iterations: 1.5 105, sample length: 1.5s) (Section 6.4)