Linking Sketch Patches by Learning Synonymous Proximity for Graphic Sketch Representation

Authors: Sicong Zang, Shikui Tu, Lei Xu

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that our method significantly improves the performance on both controllable sketch synthesis and sketch healing.
Researcher Affiliation Academia Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai, China {sczang, tushikui, leixu}@sjtu.edu.cn
Pseudocode No The paper describes the methodology with equations and textual descriptions but does not include a formally labeled 'Pseudocode' or 'Algorithm' block.
Open Source Code Yes 1The codes are in: https://github.com/CMACH508/SP-gra2seq.
Open Datasets Yes Three datasets from Quick Draw (Ha and Eck 2018) are selected for experimental comparison.
Dataset Splits Yes Each category contains 70K training, 2.5K valid and 2.5K test samples (1K = 1000).
Hardware Specification No The paper does not provide specific details about the hardware used for experiments, such as CPU or GPU models, or memory specifications.
Software Dependencies No The paper mentions using 'Adam optimizer' and 'Re LU activation function' but does not specify software dependencies with version numbers (e.g., 'PyTorch 1.9', 'Python 3.8').
Experiment Setup Yes When training an SP-gra2seq, the patch number M, the mini-batch size N, the learning rate η for updating clustering centroids and the weight λ in the objective are fixed at 20, 256, 0.05 and 0.25, respectively. The numbers of cluster centroids K are 30, 50 and 50 for three datasets, respectively. We employ Adam optimizer for the network learning with the parameters β1 = 0.9, β2 = 0.999 and ϵ = 10 8. And the learning rate starts from 10 3 with a decay rate of 0.95 for each training epoch.