Rethinking the Bottom-Up Framework for Query-Based Video Localization
Authors: Long Chen, Chujie Lu, Siliang Tang, Jun Xiao, Dong Zhang, Chilie Tan, Xiaolin Li10551-10558
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on two challenging query-based video localization tasks (natural language video localization and video relocalization), involving four challenging benchmarks (TACo S, Charades-STA, Activity Net Captions, and Activity VRL), have shown that GDP surpasses the state-of-the-art top-down models. |
| Researcher Affiliation | Collaboration | Long Chen,1 Chujie Lu,1 Siliang Tang,1 Jun Xiao,1 Dong Zhang,2 Chilie Tan,3 Xiaolin Li3,4 1DCD Lab, College of Computer Science, Zhejiang University, 2Nanjing University of Science and Technology 3Tongdun Technology 4University of Florida |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | No | The paper does not provide any concrete access information (e.g., specific repository link, explicit code release statement) for source code. |
| Open Datasets | Yes | TACo S (Regneri et al. 2013): It consists of 127 videos and 17,344 text-to-clip pairs. [...] Charades STA (Gao et al. 2017): It consists of 12,408 text-to-clip pairs for training, and 3,720 pairs for test. [...] Activity Net Captions (Krishna et al. 2017): It is the largest NLVL benchmark with much more diverse context. [...] Activity Net-VRL (Feng et al. 2018) |
| Dataset Splits | Yes | TACo S (Regneri et al. 2013): ...We used the standard split as (Gao et al. 2017), i.e., 50% for training, 25% for validation, and 25% for test. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper mentions software components like 'Adam optimizer' but does not specify version numbers for these or other key software dependencies (e.g., deep learning frameworks, Python versions). |
| Experiment Setup | Yes | The dimension of all intermediate layers was set to 128. The node number N1, N2 and N3 were set to 10. We trained the whole network from scratch with Adam optimizer for 100 epochs. The initial learning rate was set to 0.0001 and it was divided by 10 when the loss arrives on plateaus. The batch size of all experiments was set to 16, and the dropout rate was 0.5. |