Describing and learning of related parts based on latent structural model in big data

Liu, Lei; Zhang, Huigang; Zhou, Jun

doi:10.1016/j.neucom.2014.12.120

Cited by 5 publications

(4 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The first dictionary is trained on p= [1,2,3,4,5,10,15,20,25,30,35,40,45,50,55] and p= [1,2,3,4,5,10,15,20,25,26] samples per category on the Extended YaleB and AR Face datasets, respectively. We consider that in this exhaustive way, the superiority of our method could be fully reflected.…”

Section: Results On Parameter Pmentioning

confidence: 99%

“…We evaluate our approach with different q= [1,2,3,4,5,10,15,20,25,30] per person on both the Extended YaleB and AR Face datasets. The results are shown in Figure 5.…”

Section: Results On Parameter Qmentioning

confidence: 99%

“…The number of the training images is t= [1,2,3,4,5,10,15,20,25,30,35,40,45,50,55] and t= [1,2,3,4,5,10,15,20,25] per category on the Extended YaleB and AR Face datasets, we can draw two conclusions from Figure 6, (i) the classification rate first rises to the peak rapidly and then tends to be stable as t increasing; (ii) there is a small impact to classification accuracy when changing p. [43] 82.20 92.10 -DL-COPAR [42] 86.47 ± 0.69 -31.11 FDDL [50] 90.01 ± 0.69 -42.48 LLC (70 local bases) [43] 90.70 96.70 -DBDL [3] 91.09 ± 0.59 -1.07 JBDC [2] 92.14 ± 0.52 -1.02 K-SVD (15 per person) [1] 93.10 98.00 -SupGraphDL-L [53] 93.44 --D-KSVD (15 per person) [57] 94.10 98.00 -LC-KSVD1 [20] 94.50 98.30 0.52 LC-KSVD2 [20] 95.00 98.80 0.49 Multi-Block Alternating Proximal [4] 95.12 --ITDL [34] 95.39 --VAE + GAN [28] 96.4 --EasyDL [35] 96.22 --LC-KSVD2 (A-15) [20] 96.70 99.00 -SRC (all training samples) [45] 97.20 99.00 20.78 CSDL-SRC(power) [24] 98.28 ± 0.57 --RRC L1 (300)…”

Section: Results On Parameter Tmentioning

confidence: 99%

“…99.87 ± 0.19 1.63 [20] 28.10 32.95 --SRC [45] 27.86 33.33 --Griffin [17] 28.30 34.10 ± 0.20 --LC-KSVD2 [20] 28.90 34.32 --Graph-matching [9] -38.10 ± 0.60 --Local NBNN [29] 33.50 ± 0.90 40.10 ± 0.10 --Latent Structural [25] 36.34 42.01 --ScSPM [49] 27.73 ± 0.51 34 imation to achieve O(ε 3 ) on imperfect atoms. In addition, we also report the computation time (ms) for classification on several methods.…”

Section: Ddlcn (A-5)mentioning

confidence: 99%

See 3 more Smart Citations

Deep Micro-Dictionary Learning and Coding Network

Tang

Wei²,

Wei

et al. 2019

2019 IEEE Winter Conference on Applications of Computer Vision (WACV)

View full text Add to dashboard Cite

In this paper, we propose a novel Deep Micro-Dictionary Learning and Coding Network (DDLCN). DDLCN has most of the standard deep learning layers (pooling, fully, connected, input/output, etc.) but the main difference is that the fundamental convolutional layers are replaced by novel compound dictionary learning and coding layers. The dictionary learning layer learns an over-complete dictionary for the input training data. At the deep coding layer, a locality constraint is added to guarantee that the activated dictionary bases are close to each other. Next, the activated dictionary atoms are assembled together and passed to the next compound dictionary learning and coding layers. In this way, the activated atoms in the first layer can be represented by the deeper atoms in the second dictionary. Intuitively, the second dictionary is designed to learn the finegrained components which are shared among the input dictionary atoms. In this way, a more informative and discriminative low-level representation of the dictionary atoms can be obtained. We empirically compare the proposed DDLCN with several dictionary learning methods and deep learning architectures. The experimental results on four popular benchmark datasets demonstrate that the proposed DDLCN achieves competitive results compared with state-of-the-art approaches.ing methods focus on extracting semantic features via a deep network. So far most studies in dictionary learning employ a shallow (single layer) architecture, e.g., currently popular dictionary learning techniques are K-SVD [1], Discriminative K-SVD (D-KSVD) [57] and Label Consistent K-SVD (LC-KSVD) [20] which decompose the training data into a dense basis and sparse coefficients. In addition, both Local Coordinate Coding (LCC) [55, 54] and arXiv:1809.04185v2 [cs.CV]

show abstract

Section: Results On Parameter Pmentioning

confidence: 99%

“…We evaluate our approach with different q= [1,2,3,4,5,10,15,20,25,30] per person on both the Extended YaleB and AR Face datasets. The results are shown in Figure 5.…”

Section: Results On Parameter Qmentioning

confidence: 99%