Deep Attribute Guided Representation for Heterogeneous Face Recognition
Authors: Decheng Liu, Nannan Wang, Chunlei Peng, Jie Li, Xinbo Gao
IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on multiple heterogeneous scenarios (composite sketches, resident ID cards) demonstrate that the proposed method achieves superior performances compared with state-of-the-art methods. |
| Researcher Affiliation | Academia | 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 Engineering, Xidian University, Xi an 710071, China School of Cyber Engineering, Xidian University, Xi an 710071, China |
| Pseudocode | No | The paper describes the method and network architecture but does not contain a dedicated 'Pseudocode' or 'Algorithm' block. |
| Open Source Code | No | The paper does not explicitly state that the source code for their methodology is publicly available, nor does it provide a link to a code repository. |
| Open Datasets | Yes | Extended PRIP Database (E-PRIP) contains 123 subjects, with photos from the AR database [Wang and Tang, 2009] and composite sketches created by FACES software. PRIP Viewed Software-Generated Composite Database (PRIPVSGC) also contains 123 subjects, with photos from the AR dataset and composite sketches created by Identi-Kit software. NJU-ID database [Huo et al., 2017] contains 256 persons. ... We pre-train this multi attributes evaluation network with Celeb A dataset [Liu et al., 2015] which contains 202,599 face images, each with 40 binary attributes labeled. ... To get a better initial parameter, we pre-train the proposed network with the Image Net database. |
| Dataset Splits | Yes | With the same protocol in [Liu et al., 2015], about 80% of images are used to fine-tune, 10% of images are used as validation data, and the rest of images are used as testing data. |
| Hardware Specification | Yes | The proposed deep attribute guided HFR method related experiments conducted on Titan X GPU. And the other experiments are conducted on an Intel Core i7-4790 3.60GHz PC under MATLAB R2014a environment. |
| Software Dependencies | Yes | And the other experiments are conducted on an Intel Core i7-4790 3.60GHz PC under MATLAB R2014a environment. |
| Experiment Setup | Yes | The input size of the network is 224 224 and the first convolution layer creates 64 outputs with filter size of 7 7 shown in Figure 1. ... We build four basic blocks on the basis of [He et al., 2016], where four kinds of convolutional layers separately generate 64, 128, 256 and 512 feature maps with filter size of 3 3 . ... We train the CNN using stochastic gradient descent and Ada Grad[Duchi et al., 2011]. We start with a small learning rate of 0.0005 and reduce the learning rate by 0.1 every 5 epochs. We use a weight decay of 0.0005 and a momentum of 0.9. |