LinkNet: Relational Embedding for Scene Graph

Authors: Sanghyun Woo, Dahun Kim, Donghyeon Cho, In So Kweon

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We validate our final model, Link Net, through extensive ablation studies, demonstrating its efficacy in scene graph generation.
Researcher Affiliation Academia Sanghyun Woo* EE, KAIST Daejeon, Korea shwoo93@kaist.ac.kr Dahun Kim* EE, KAIST Daejeon, Korea mcahny@kaist.ac.kr Donghyeon Cho EE, KAIST Daejeon, Korea cdh12242@gmail.com In So Kweon EE, KAIST Daejeon, Korea iskweon@kaist.ac.kr
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks. It describes the model architecture and mathematical formulations.
Open Source Code No The paper does not provide concrete access to source code for the methodology described in this paper.
Open Datasets Yes We conduct experiments on Visual Genome benchmark [16].
Dataset Splits No The paper mentions using the Visual Genome benchmark but does not explicitly provide the training/test/validation dataset splits or reference a specific standard split with enough detail within the paper itself.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiment.
Experiment Setup Yes In practice, we found setting hyper-parameter r as 2 produces best result from our experiment. By default, we set λ1 and λ2 as 1, and thus all the terms are equally weighted.