Actionness Inconsistency-Guided Contrastive Learning for Weakly-Supervised Temporal Action Localization
Authors: Zhilin Li, Zilei Wang, Qinying Liu
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on THUMOS14, Activity Net v1.2, and Activity Net v1.3 datasets, and the results show the effectiveness of AICL with state-of-the-art performance. |
| Researcher Affiliation | Academia | University of Science and Technology of China, Hefei, China lizhilin@mail.ustc.edu.cn, zlwang@ustc.edu.cn, lydyc@mail.ustc.edu.cn |
| Pseudocode | No | The paper describes the proposed method using textual explanations and mathematical equations, but it does not include a distinct section or figure labeled 'Pseudocode' or 'Algorithm' with structured steps. |
| Open Source Code | Yes | Our code is available at https://github.com/ lizhilin-ustc/AAAI2023-AICL. |
| Open Datasets | Yes | We conduct extensive experiments on THUMOS14, Activity Net v1.2, and Activity Net v1.3 datasets, and the results show the effectiveness of AICL with state-of-the-art performance. |
| Dataset Splits | Yes | THUMOS14: 200 validation videos to train our framework and 213 test videos for testing. Activity Net1.2: 4,819 training videos, 2,383 validation videos, and 2,489 test videos of 100 action classes. Activity Net1.3 contains 10,024 training videos and 4,926 validation videos from 200 action categories. Following previous work (Gao, Chen, and Xu 2022; Li et al. 2022; Huang, Wang, and Li 2022), we train on the training set and test on the validation set. |
| Hardware Specification | No | The Acknowledgements section mentions 'GPU cluster built by MCC Lab of Information Science and Technology Institution, USTC', but no specific hardware models (e.g., GPU type, CPU, memory) are provided for the experiments. |
| Software Dependencies | No | The paper mentions using 'I3D (Carreira and Zisserman 2017) model pretrained on Kinetics' and describes network layers like 'convolution layer and RELU activations', but it does not specify any software libraries (e.g., PyTorch, TensorFlow) or their version numbers, nor the Python version used. |
| Experiment Setup | Yes | We set noise tolerance q = 0.7 for both datasets, and use instance selection parameters k = T/8 for THUMOS14. We set γ to be 1/3 for the Ascore in THUMOS14. We set K = T/20 to choose the number of inconsistent action and background segments. We set γ1 = 0.01 for THUMOS14. |