CT-Net: Channel Tensorization Network for Video Classification

Authors: Kunchang Li, Xianhang Li, Yali Wang, Jun Wang, Yu Qiao

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments are conducted on several challenging video benchmarks, e.g., Kinetics-400, Something-Something V1 and V2.
Researcher Affiliation Academia 1Guangdong-Hong Kong-Macao Joint Laboratory of Human-Machine Intelligence-Synergy Systems, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518055, China 2University of Chinese Academy of Sciences 3SIAT Branch, Shenzhen Institute of Artificial Intelligence and Robotics for Society 4University of Central Florida
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks. Figure 4 provides a diagram of the Tensor Excitation mechanism, but it is not pseudocode.
Open Source Code No The paper does not provide any explicit statements about the release of source code or include any links to code repositories.
Open Datasets Yes We conduct experiments on three large video benchmarks: Kinetics-400 (Carreira & Zisserman, 2017), Something-Something V1 and V2 (Goyal et al., 2017b)... To verify the generation ability of our CT-Net on smaller datasets, we conduct transfer learning experiments from Kinetics400 to UCF101 (Soomro et al., 2012) and HMDB-51 (Kuehne et al., 2011).
Dataset Splits Yes We conduct experiments on three large video benchmarks: Kinetics-400 (Carreira & Zisserman, 2017), Something-Something V1 and V2 (Goyal et al., 2017b)... We test CT-Net with 16 input frames and evaluate it over three splits and report the averaged results.
Hardware Specification No The paper does not specify the hardware used for running experiments, such as particular GPU or CPU models.
Software Dependencies No The paper mentions optimization techniques (SGD with momentum, cosine learning rate schedule) and references other models/methods (ResNet, Non-local), but does not provide specific version numbers for software dependencies or libraries like PyTorch or TensorFlow.
Experiment Setup Yes For kinetics, the batch, total epochs, initial learning rate, dropout and weight decay are set to 64, 110, 0.01, 0.5 and 1e-4 respectively. All these hyper-parameters are set to 64, 45, 0.02, 0.3 and 5e-4 respectively for Something-Something.