Exploring Domain Incremental Video Highlights Detection with the LiveFood Benchmark
Authors: Sen Pei, Shixiong Xu, Xiaojie Jin
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the effectiveness of GPE through extensive experiments. Notably, GPE surpasses popular domain incremental learning methods on Live Food, achieving significant m AP improvements on all domains. |
| Researcher Affiliation | Collaboration | Sen Pei1, Shixiong Xu2, and Xiaojie Jin1* 1 Byte Dance Inc. 2 Institute of Automation, Chinese Academy of Sciences |
| Pseudocode | No | The paper describes the model's architecture and learning process using text and mathematical formulations, but it does not include a dedicated figure, block, or section explicitly labeled "Pseudocode" or "Algorithm" in the provided text. |
| Open Source Code | Yes | The code is available at: https://github.com/ Forever Ps/Incremental VHD GPE. |
| Open Datasets | No | The paper states that the Live Food dataset was collected by the authors, but it does not provide any specific link, DOI, repository name, or formal citation for accessing this dataset publicly. It only describes its characteristics and use in the experiments. |
| Dataset Splits | Yes | Live Food contains 4928 videos for training and 261 videos for testing. We randomly split 15% of the 4928 videos for validation. |
| Hardware Specification | No | The paper describes the computational models and experimental procedures but does not provide any specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments. |
| Software Dependencies | No | The paper mentions using pre-trained models and specific architectures (e.g., ConvNeXt, ResNet-18) but does not specify any software dependencies with their version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions). |
| Experiment Setup | Yes | We randomly split 15% of the 4928 videos for validation. T1, T2, T3 and T4 consist of 3380, 854, 393, and 113 videos, respectively. ... We set k to 40 throughout experiments to strike a good balance between accuracy and efficiency. ... Consequently, we set k and γ to 40 and 5 by default. |