Deep Graphical Feature Learning for Face Sketch Synthesis

Authors: Mingrui Zhu, Nannan Wang, Xinbo Gao, Jie Li

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on public face sketch databases show that our method outperforms state-of-the-art methods, in terms of both synthesis quality and recognition ability. In this section, we report the experimental results of the proposed DGFL method quantitatively and qualitatively. We conducted our experiments on the Chinese University of Hong Kong (CUHK) face sketch database (CUFS) [Tang and Wang, 2009].
Researcher Affiliation Academia Mingrui Zhu , Nannan Wang , Xinbo Gao , Jie Li State Key Laboratory of Integrated Services Networks, School of Electronic Engineering, Xidian University, Xi an 710071, China State Key Laboratory of Integrated Services Networks, School of Telecommunications, Xidian University, Xi an 710071, China mrz.edu@gmail.com, nnwang@xidian.edu.cn, {leejie,xbgao}@mail.xidian.edu.cn
Pseudocode Yes Algorithm 1 Deep Graphical Feature Learning for Face Sketch Synthesis
Open Source Code No The paper states that
Open Datasets Yes We conducted our experiments on the Chinese University of Hong Kong (CUHK) face sketch database (CUFS) [Tang and Wang, 2009]. The CUFS database consists of face photos from three databases: the CUHK student database [Tang and Wang, 2002] (188 persons), the AR database [Martinez and Benavente, 1998] (123 persons) and the XM2VTS database [K. Messer and J. Luettin, 1999] (295 persons).
Dataset Splits Yes For the CUHK student database, 88 pairs of face photo-sketch are taken for training and the rest for testing. For the AR database, we randomly choose 80 pairs for training and the rest 43 pairs for testing. For the XM2VTS database, we randomly choose 100 pairs for training and the rest 195 pairs for testing.
Hardware Specification Yes All experiments are conducted using Python on Ubuntu 14.04 system with i7-4790 3.6G CPU and 12G NVIDIA Titan X GPU.
Software Dependencies No The paper mentions "Python on Ubuntu 14.04 system" but does not specify exact version numbers for Python itself or any other relevant software libraries (e.g., TensorFlow, PyTorch, NumPy) used for the implementation.
Experiment Setup Yes The parameters used were set as follows: the image patch size was 10, the overlap size was 5, the size of search region was 5, the number of candidate patches K was set to 10, the α was set to 0.25, the λ was set to 2.