Arcabouço Computacional para Segmentação e Restauração Digital de Artefatos em Imagens Frontais de Face

Sobiecki, André; Giraldi, Gilson A.; Neves, Luiz Antônio Pereira; Gattás, Gilka Jorge Fígaro; Thomaz, Carlos Eduardo

doi:10.7436/2012.avc.17

Cited by 1 publication

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus we can assume that in the case of FaceNet, not only DeblurGAN and FSRNet have a improved accuracy but also the mixing method. SphereFace present the largest discrepancy, at the equal error rate only the accuracy improves with DeblurGAN but the AUC scores better with [30], NeuralEnhance, DeblurGAN and [29]. For FaceNet the difference in accuracy between DeblurGAN and the resized images is 3%.…”

Section: Face Verification After Restorationmentioning

confidence: 98%

“…Since we cannot in general measure ground-truth (we don't have 'ideal' images of the faces of persons in I ), we evaluate the quality of f restor by using 'proxy' quality metrics that look for typical things found in good-quality images, like contrast, details, texture and the results of algorithms of face recognition. Most of the methods of image restoration works on any image but there there are some methods that are made only for face image restoration such as [6,30] and [29].…”

Section: Face or Image Restorationmentioning

confidence: 99%

“…where restoration SR FSRNet CNN yes [12] restoration SR Bicubic interpolation Traditional no [30] restoration deblur Beter digitalization Segmentation, inpaint yes [29] restoration deblur Face Inpaint Segmentation, inpaint yes [14] restoration deblur DeblurGAN CNN no [34] verification Eigenfaces Covariance matrix yes [2] verification Fisherfaces Linear discriminant yes analysis [1] verification local binary patterns CNN yes [39] verification LightCNN CNN yes [26] verification FaceNet CNN yes [18] verification SphereFace CNN yes indicate uniform color areas. In our experiments we next compare I result with I F SRNet and with all other methods for face restoration.…”

Section: Mixing Fsrnet With the Original Imagementioning

confidence: 99%

“…When choosing a slightly higher threshold, the number of false positives will drop dramatically increasing the overal accuracy. [30], h [29] and i mixing F SRNet with original…”

Section: Face Verificationmentioning

confidence: 99%

“…Fig. 5 Surveillance dataset: a original, b bicubic interpolation, c Smart Deblur, d Neural Enhance, e DeblurGAN, f FSRNet, g[30], h[29] and i mixing F SRNet with original…”

mentioning

confidence: 99%

See 4 more Smart Citations

Does face restoration improve face verification?

Sobiecki¹,

Dijk

Folkertsma

et al. 2021

Multimed Tools Appl

Self Cite

View full text Add to dashboard Cite

Methods for face verification works reasonably well on face images with standardized (frontal) face positions and good spatial resolution. However such methods have significant challenges on poor resolution images, poor lighting conditions and not standard (frontal) face positions. In this paper, we survey the capability of existing face restoration and verification methods, with the aim of understanding how useful face restoration methods are for face verification. We propose a qualitative and quantitative comparison benchmark, and apply it on eight methods for face restoration and six methods for face verification, on several real-world low-quality images from a surveillance context, and outline observed advantages and limitations. Experiments shows that each restoration method can affect each face verification method differently, with fewer than the half of face restoration methods helping face verification. Interestingly, some face restoration methods with less good qualitative evaluation helped face verification the most. Experiments also show that face verification works less good if the resolution decreases.

show abstract

Section: Face Verification After Restorationmentioning

confidence: 98%

Section: Face or Image Restorationmentioning

confidence: 99%

Section: Mixing Fsrnet With the Original Imagementioning

confidence: 99%

“…When choosing a slightly higher threshold, the number of false positives will drop dramatically increasing the overal accuracy. [30], h [29] and i mixing F SRNet with original…”

Section: Face Verificationmentioning

confidence: 99%

“…Fig. 5 Surveillance dataset: a original, b bicubic interpolation, c Smart Deblur, d Neural Enhance, e DeblurGAN, f FSRNet, g[30], h[29] and i mixing F SRNet with original…”

mentioning

confidence: 99%

See 3 more Smart Citations