Feature Augmented Memory with Global Attention Network for VideoQA

Authors: Jiayin Cai, Chun Yuan, Cheng Shi, Lei Li, Yangyang Cheng, Ying Shan

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that our approach achieves state-of-the-art performance on Video QA benchmark datasets.
Researcher Affiliation Collaboration Jiayin Cai1 , Chun Yuan2 , Cheng Shi1 , Lei Li1 , Yangyang Cheng1 , Ying Shan3 1Department of Computer Science and Technology, Tsinghua University 2Tsinghua Shenzhen International Graduate School 3ARC, Tencent PCG
Pseudocode No The paper describes its models and processes using text and mathematical equations, but it does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statements about releasing source code or links to a code repository.
Open Datasets Yes We evaluate our model on three benchmark Video QA dataset TGIF-QA, MSVD-QA and MSRVTT-QA respectively... TGIF-QA Dataset [Jang et al., 2017]... MSVD-QA and MSRVTT-QA Dataset were proposed by [Xu et al., 2017] based on MSVD [L and B, 2011] and MSVTT [Xu et al., 2016] video sets respectively.
Dataset Splits No The paper states: "The batch size is set as 32. The train epoch is set as 30." It mentions datasets like TGIF-QA, MSVD-QA, and MSRVTT-QA and refers to 'test QA-pair numbers', but it does not explicitly specify the proportions or counts for training, validation, and test splits needed for reproduction.
Hardware Specification No The paper does not explicitly state the specific hardware (e.g., GPU/CPU models, memory) used for running the experiments. It only mentions general training parameters like batch size and epochs.
Software Dependencies No The paper mentions using specific models like "C3D [Tran et al., 2015] network", "pretrained Res Net [He et al., 2016] or VGG [Simonyan and Zisserman, 2015] network", and "pre-trained Glo Ve 300-D feature", and that "the optimization algorithm is Adam". However, it does not provide specific version numbers for any software libraries, frameworks, or languages used.
Experiment Setup Yes In our experiments, the optimization algorithm is Adam. The batch size is set as 32. The train epoch is set as 30. In addition, gradient clipping, weight normalization are employed in training.