Image recolorization enhances the visual perception of an image for design and artistic purposes. In this work, we present a deep neural network, referred to as PaletteNet, which recolors an image according to a given target color palette that is useful to express the color concept of an image. PaletteNet takes two inputs: a source image to be recolored and a target palette. PaletteNet is then designed to change the color concept of a source image so that the palette of the output image is close to the target palette. To train PaletteNet, the proposed multi-task loss is composed of Euclidean loss and adversarial loss. The experimental results show that the proposed method outperforms the existing recolorization methods. Human experts with a commercial software take on average 18 minutes to recolor an image, while PaletteNet automatically recolors plausible results in less than a second.
{"title":"PaletteNet: Image Recolorization with Given Color Palette","authors":"Junho Cho, Sangdoo Yun, Kyoung-Ok Lee, J. Choi","doi":"10.1109/CVPRW.2017.143","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.143","url":null,"abstract":"Image recolorization enhances the visual perception of an image for design and artistic purposes. In this work, we present a deep neural network, referred to as PaletteNet, which recolors an image according to a given target color palette that is useful to express the color concept of an image. PaletteNet takes two inputs: a source image to be recolored and a target palette. PaletteNet is then designed to change the color concept of a source image so that the palette of the output image is close to the target palette. To train PaletteNet, the proposed multi-task loss is composed of Euclidean loss and adversarial loss. The experimental results show that the proposed method outperforms the existing recolorization methods. Human experts with a commercial software take on average 18 minutes to recolor an image, while PaletteNet automatically recolors plausible results in less than a second.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"23 1","pages":"1058-1066"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78541979","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
R. Timofte, E. Agustsson, L. Gool, Ming-Hsuan Yang, Lei Zhang, Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, Kyoung Mu Lee, Xintao Wang, Yapeng Tian, K. Yu, Yulun Zhang, Shixiang Wu, Chao Dong, Liang Lin, Y. Qiao, Chen Change Loy, Woong Bae, J. Yoo, Yoseob Han, J. C. Ye, Jae-Seok Choi, Munchurl Kim, Yuchen Fan, Jiahui Yu, Wei Han, Ding Liu, Haichao Yu, Zhangyang Wang, Humphrey Shi, Xinchao Wang, Thomas S. Huang, Yunjin Chen, K. Zhang, W. Zuo, Zhimin Tang, Linkai Luo, Shaohui Li, Min Fu, Lei Cao, Wen Heng, Giang Bui, Truc Le, Y. Duan, D. Tao, Ruxin Wang, Xu Lin, Jianxin Pang, Jinchang Xu, Yu Zhao, Xiangyu Xu, Jin-shan Pan, Deqing Sun, Yujin Zhang, Xibin Song, Yuchao Dai, Xueying Qin, X. Huynh, Tiantong Guo, Hojjat Seyed Mousavi, T. Vu, V. Monga, Cristóvão Cruz, K. Egiazarian, V. Katkovnik, Rakesh Mehta, A. Jain, Abhinav Agarwalla, C. Praveen, Ruofan Zhou, Hongdiao Wen, Chen Zhu, Zhiqiang Xia, Zhengtao Wang, Qi Guo
This paper reviews the first challenge on single image super-resolution (restoration of rich details in an low resolution image) with focus on proposed solutions and results. A new DIVerse 2K resolution image dataset (DIV2K) was employed. The challenge had 6 competitions divided into 2 tracks with 3 magnification factors each. Track 1 employed the standard bicubic downscaling setup, while Track 2 had unknown downscaling operators (blur kernel and decimation) but learnable through low and high res train images. Each competition had ∽100 registered participants and 20 teams competed in the final testing phase. They gauge the state-of-the-art in single image super-resolution.
{"title":"NTIRE 2017 Challenge on Single Image Super-Resolution: Methods and Results","authors":"R. Timofte, E. Agustsson, L. Gool, Ming-Hsuan Yang, Lei Zhang, Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, Kyoung Mu Lee, Xintao Wang, Yapeng Tian, K. Yu, Yulun Zhang, Shixiang Wu, Chao Dong, Liang Lin, Y. Qiao, Chen Change Loy, Woong Bae, J. Yoo, Yoseob Han, J. C. Ye, Jae-Seok Choi, Munchurl Kim, Yuchen Fan, Jiahui Yu, Wei Han, Ding Liu, Haichao Yu, Zhangyang Wang, Humphrey Shi, Xinchao Wang, Thomas S. Huang, Yunjin Chen, K. Zhang, W. Zuo, Zhimin Tang, Linkai Luo, Shaohui Li, Min Fu, Lei Cao, Wen Heng, Giang Bui, Truc Le, Y. Duan, D. Tao, Ruxin Wang, Xu Lin, Jianxin Pang, Jinchang Xu, Yu Zhao, Xiangyu Xu, Jin-shan Pan, Deqing Sun, Yujin Zhang, Xibin Song, Yuchao Dai, Xueying Qin, X. Huynh, Tiantong Guo, Hojjat Seyed Mousavi, T. Vu, V. Monga, Cristóvão Cruz, K. Egiazarian, V. Katkovnik, Rakesh Mehta, A. Jain, Abhinav Agarwalla, C. Praveen, Ruofan Zhou, Hongdiao Wen, Chen Zhu, Zhiqiang Xia, Zhengtao Wang, Qi Guo","doi":"10.1109/CVPRW.2017.149","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.149","url":null,"abstract":"This paper reviews the first challenge on single image super-resolution (restoration of rich details in an low resolution image) with focus on proposed solutions and results. A new DIVerse 2K resolution image dataset (DIV2K) was employed. The challenge had 6 competitions divided into 2 tracks with 3 magnification factors each. Track 1 employed the standard bicubic downscaling setup, while Track 2 had unknown downscaling operators (blur kernel and decimation) but learnable through low and high res train images. Each competition had ∽100 registered participants and 20 teams competed in the final testing phase. They gauge the state-of-the-art in single image super-resolution.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"4 1","pages":"1110-1121"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87866449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A major challenge in visual highway traffic analytics is to disaggregate individual vehicles from clusters formed in dense traffic conditions. Here we introduce a data driven 3D generative reasoning method to tackle this segmentation problem. The method is comprised of offline (learning) and online (inference) stages. In the offline stage, we fit a mixture model for the prior distribution of vehicle dimensions to labelled data. Given camera intrinsic parameters and height, we use a parallelism method to estimate highway lane structure and camera tilt to project 3D models to the image. In the online stage, foreground vehicle cluster segments are extracted using motion and background subtraction. For each segment, we use a data-driven MCMC method to estimate the vehicles configuration and dimensions that provide the most likely account of the observed foreground pixels. We evaluate the method on two highway datasets and demonstrate a substantial improvement on the state of the art.
{"title":"Slot Cars: 3D Modelling for Improved Visual Traffic Analytics","authors":"Eduardo R. Corral-Soto, J. Elder","doi":"10.1109/CVPRW.2017.123","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.123","url":null,"abstract":"A major challenge in visual highway traffic analytics is to disaggregate individual vehicles from clusters formed in dense traffic conditions. Here we introduce a data driven 3D generative reasoning method to tackle this segmentation problem. The method is comprised of offline (learning) and online (inference) stages. In the offline stage, we fit a mixture model for the prior distribution of vehicle dimensions to labelled data. Given camera intrinsic parameters and height, we use a parallelism method to estimate highway lane structure and camera tilt to project 3D models to the image. In the online stage, foreground vehicle cluster segments are extracted using motion and background subtraction. For each segment, we use a data-driven MCMC method to estimate the vehicles configuration and dimensions that provide the most likely account of the observed foreground pixels. We evaluate the method on two highway datasets and demonstrate a substantial improvement on the state of the art.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"1 1","pages":"889-897"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87037028","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Philip Saponaro, Wayne Treible, Abhishek Kolagunda, Timothy Chaya, J. Caplan, C. Kambhamettu, R. Wisser
High-speed confocal microscopy has shown great promise to yield insights into plant-fungal interactions by allowing for large volumes of leaf tissue to be imaged at high magnification. Currently, segmentation is performed either manually, which is infeasible for large amounts of data, or by developing separate algorithms to extract individual features within the image data. In this work, we propose the use of a single deep convolutional neural network architecture dubbed DeepXScope for automatically segmenting hyphal networks of the fungal pathogen and cell boundaries and stomata of the host plant. DeepXScope is trained on manually annotated images created for each of these structures. We describe experiments that show each individual structure can be accurately extracted automatically using DeepXScope. We anticipate that plant scientists will be able to use this network to automatically extract multiple structures of interest, and we plan to release our tool to the community1.
{"title":"DeepXScope: Segmenting Microscopy Images with a Deep Neural Network","authors":"Philip Saponaro, Wayne Treible, Abhishek Kolagunda, Timothy Chaya, J. Caplan, C. Kambhamettu, R. Wisser","doi":"10.1109/CVPRW.2017.117","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.117","url":null,"abstract":"High-speed confocal microscopy has shown great promise to yield insights into plant-fungal interactions by allowing for large volumes of leaf tissue to be imaged at high magnification. Currently, segmentation is performed either manually, which is infeasible for large amounts of data, or by developing separate algorithms to extract individual features within the image data. In this work, we propose the use of a single deep convolutional neural network architecture dubbed DeepXScope for automatically segmenting hyphal networks of the fungal pathogen and cell boundaries and stomata of the host plant. DeepXScope is trained on manually annotated images created for each of these structures. We describe experiments that show each individual structure can be accurately extracted automatically using DeepXScope. We anticipate that plant scientists will be able to use this network to automatically extract multiple structures of interest, and we plan to release our tool to the community1.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"2014 1","pages":"843-850"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88101538","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In the last years, many computer vision algorithms have been developed for X-ray testing tasks. Some of them deal with baggage inspection, in which the aim is to detect automatically target objects. The progress in automated baggage inspection, however, is modest and very limited compared to what is needed because X-ray screening systems are still being manipulated by human inspectors. In this work, we present an X-ray imaging model that can separate foreground from background in baggage screening. The model can be used in two main tasks: i) Simulation of new X-ray images, where simulated images can be used in training programs for human inspectors, or can be used to enhance datasets for computer vision algorithms. ii) Detection of (threat) objects, where new algorithms can be employed to perform automated baggage inspection or to aid an user in the inspection task showing potential threats. In our model, rather than a multiplication of foreground and background, that is typically used in X-ray imaging, we propose the addition of logarithmic images. This allows the use of linear strategies to superimpose images of threat objects onto X-ray images and the use of sparse representations in order to segment target objects. In our experiments, we simulate new X-ray images of handguns, shuriken and razor blades, in which it is impossible to distinguish simulated and real X-ray images. In addition, we show in our experiments the effective detection of shuriken, razor blades and handguns using the proposed algorithm outperforming some alternative state-of- the-art techniques.
{"title":"A Logarithmic X-Ray Imaging Model for Baggage Inspection: Simulation and Object Detection","authors":"D. Mery, A. Katsaggelos","doi":"10.1109/CVPRW.2017.37","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.37","url":null,"abstract":"In the last years, many computer vision algorithms have been developed for X-ray testing tasks. Some of them deal with baggage inspection, in which the aim is to detect automatically target objects. The progress in automated baggage inspection, however, is modest and very limited compared to what is needed because X-ray screening systems are still being manipulated by human inspectors. In this work, we present an X-ray imaging model that can separate foreground from background in baggage screening. The model can be used in two main tasks: i) Simulation of new X-ray images, where simulated images can be used in training programs for human inspectors, or can be used to enhance datasets for computer vision algorithms. ii) Detection of (threat) objects, where new algorithms can be employed to perform automated baggage inspection or to aid an user in the inspection task showing potential threats. In our model, rather than a multiplication of foreground and background, that is typically used in X-ray imaging, we propose the addition of logarithmic images. This allows the use of linear strategies to superimpose images of threat objects onto X-ray images and the use of sparse representations in order to segment target objects. In our experiments, we simulate new X-ray images of handguns, shuriken and razor blades, in which it is impossible to distinguish simulated and real X-ray images. In addition, we show in our experiments the effective detection of shuriken, razor blades and handguns using the proposed algorithm outperforming some alternative state-of- the-art techniques.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"73 1","pages":"251-259"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83978837","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Recent work has demonstrated the emergence of semantic object-part detectors in activation patterns of convolutional neural networks (CNNs), but did not account for the distributed multi-layer neural activations in such networks. In this work, we propose a novel method to extract distributed patterns of activations from a CNN and show that such patterns correspond to high-level visual attributes. We propose an unsupervised learning module that sits above a pre-trained CNN and learns distributed activation patterns of the network. We utilize elastic non-negative matrix factorization to analyze the responses of a pretrained CNN to an input image and extract salient image regions. The corresponding patterns of neural activations for the extracted salient regions are then clustered via unsupervised deep embedding for clustering (DEC) framework. We demonstrate that these distributed activations contain high-level image features that could be explicitly used for image classification.
{"title":"Explaining Distributed Neural Activations via Unsupervised Learning","authors":"Soheil Kolouri, Charles E. Martin, Heiko Hoffmann","doi":"10.1109/CVPRW.2017.213","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.213","url":null,"abstract":"Recent work has demonstrated the emergence of semantic object-part detectors in activation patterns of convolutional neural networks (CNNs), but did not account for the distributed multi-layer neural activations in such networks. In this work, we propose a novel method to extract distributed patterns of activations from a CNN and show that such patterns correspond to high-level visual attributes. We propose an unsupervised learning module that sits above a pre-trained CNN and learns distributed activation patterns of the network. We utilize elastic non-negative matrix factorization to analyze the responses of a pretrained CNN to an input image and extract salient image regions. The corresponding patterns of neural activations for the extracted salient regions are then clustered via unsupervised deep embedding for clustering (DEC) framework. We demonstrate that these distributed activations contain high-level image features that could be explicitly used for image classification.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"58 1","pages":"1670-1678"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80624811","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Bor-Chun Chen, P. Ghosh, Vlad I. Morariu, L. Davis
Image content or metadata editing software availability and ease of use has resulted in a high demand for automatic image tamper detection algorithms. Most previous work has focused on detection of tampered image content, whereas we develop techniques to detect metadata tampering in outdoor images using sun altitude angle and other meteorological information like temperature, humidity and weather, which can be observed in most outdoor image scenes. To train and evaluate our technique, we create a large dataset of outdoor images labeled with sun altitude angle and other meteorological data (AMOS+M2), which to our knowledge, is the largest publicly available dataset of its kind. Using this dataset, we train separate regression models for sun altitude angle, temperature and humidity and a classification model for weather to detect any discrepancy between image content and its metadata. Finally, a joint multi-task network for these four features shows a relative improvement of 15.5% compared to each of them individually. We include a detailed analysis for using these networks to detect various types of modification to location and time information in image metadata.
{"title":"Detection of Metadata Tampering Through Discrepancy Between Image Content and Metadata Using Multi-task Deep Learning","authors":"Bor-Chun Chen, P. Ghosh, Vlad I. Morariu, L. Davis","doi":"10.1109/CVPRW.2017.234","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.234","url":null,"abstract":"Image content or metadata editing software availability and ease of use has resulted in a high demand for automatic image tamper detection algorithms. Most previous work has focused on detection of tampered image content, whereas we develop techniques to detect metadata tampering in outdoor images using sun altitude angle and other meteorological information like temperature, humidity and weather, which can be observed in most outdoor image scenes. To train and evaluate our technique, we create a large dataset of outdoor images labeled with sun altitude angle and other meteorological data (AMOS+M2), which to our knowledge, is the largest publicly available dataset of its kind. Using this dataset, we train separate regression models for sun altitude angle, temperature and humidity and a classification model for weather to detect any discrepancy between image content and its metadata. Finally, a joint multi-task network for these four features shows a relative improvement of 15.5% compared to each of them individually. We include a detailed analysis for using these networks to detect various types of modification to location and time information in image metadata.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"17 1","pages":"1872-1880"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91141718","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Convolutional neural networks (CNNs) excel in various computer vision related tasks but are extremely computationally intensive and power hungry to run on mobile and embedded devices. Recent pruning techniques can reduce the computation and memory requirements of CNNs, but a costly retraining step is needed to restore the classification accuracy of the pruned model. In this paper, we present evidence that when only a subset of the classes need to be classified, we could prune a model and achieve reasonable classification accuracy without retraining. The resulting specialist model will require less energy and time to run than the original full model. To compensate for the pruning, we take advantage of the redundancy among filters and class-specific features. We show that even simple methods such as replacing channels with mean or with the most correlated channel can boost the accuracy of the pruned model to reasonable levels.
{"title":"Pruning ConvNets Online for Efficient Specialist Models","authors":"Jia Guo, M. Potkonjak","doi":"10.1109/CVPRW.2017.58","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.58","url":null,"abstract":"Convolutional neural networks (CNNs) excel in various computer vision related tasks but are extremely computationally intensive and power hungry to run on mobile and embedded devices. Recent pruning techniques can reduce the computation and memory requirements of CNNs, but a costly retraining step is needed to restore the classification accuracy of the pruned model. In this paper, we present evidence that when only a subset of the classes need to be classified, we could prune a model and achieve reasonable classification accuracy without retraining. The resulting specialist model will require less energy and time to run than the original full model. To compensate for the pruning, we take advantage of the redundancy among filters and class-specific features. We show that even simple methods such as replacing channels with mean or with the most correlated channel can boost the accuracy of the pruned model to reasonable levels.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"17 2 1","pages":"430-437"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91206930","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
G. Hsu, Yi-Tseng Cheng, Choon-Ching Ng, Moi Hoon Yap
We propose the Component Bio-Inspired Feature (CBIF) with a moving segmentation scheme for age estimation. The CBIF defines a superset for the commonly used Bio-Inspired Feature (BIF) with more parameters and flexibility in settings, resulting in features with abundant characteristics. An in-depth study is performed for the determination of the parameters good for capturing age-related traits. The moving segmentation is proposed to better determine the age boundaries good for age grouping, and improve the overall performance. The proposed approach is evaluated on two common benchmarks, FG-NET and MORPH databases, and compared with contemporary approaches to demonstrate its efficacy.
{"title":"Component Biologically Inspired Features with Moving Segmentation for Age Estimation","authors":"G. Hsu, Yi-Tseng Cheng, Choon-Ching Ng, Moi Hoon Yap","doi":"10.1109/CVPRW.2017.81","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.81","url":null,"abstract":"We propose the Component Bio-Inspired Feature (CBIF) with a moving segmentation scheme for age estimation. The CBIF defines a superset for the commonly used Bio-Inspired Feature (BIF) with more parameters and flexibility in settings, resulting in features with abundant characteristics. An in-depth study is performed for the determination of the parameters good for capturing age-related traits. The moving segmentation is proposed to better determine the age boundaries good for age grouping, and improve the overall performance. The proposed approach is evaluated on two common benchmarks, FG-NET and MORPH databases, and compared with contemporary approaches to demonstrate its efficacy.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"36 1","pages":"540-547"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76609977","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ramachandra Raghavendra, K. Raja, S. Venkatesh, C. Busch
Face biometrics is widely used in various applications including border control and facilitating the verification of travellers' identity claim with respect to his electronic passport (ePass). As in most countries, passports are issued to a citizen based on the submitted photo which allows the applicant to provide a morphed face photo to conceal his identity during the application process. In this work, we propose a novel approach leveraging the transferable features from a pre-trained Deep Convolutional Neural Networks (D-CNN) to detect both digital and print-scanned morphed face image. Thus, the proposed approach is based on the feature level fusion of the first fully connected layers of two D-CNN (VGG19 and AlexNet) that are specifically fine-tuned using the morphed face image database. The proposed method is extensively evaluated on the newly constructed database with both digital and print-scanned morphed face images corresponding to bona fide and morphed data reflecting a real-life scenario. The obtained results consistently demonstrate improved detection performance of the proposed scheme over previously proposed methods on both the digital and the print-scanned morphed face image database.
{"title":"Transferable Deep-CNN Features for Detecting Digital and Print-Scanned Morphed Face Images","authors":"Ramachandra Raghavendra, K. Raja, S. Venkatesh, C. Busch","doi":"10.1109/CVPRW.2017.228","DOIUrl":"https://doi.org/10.1109/CVPRW.2017.228","url":null,"abstract":"Face biometrics is widely used in various applications including border control and facilitating the verification of travellers' identity claim with respect to his electronic passport (ePass). As in most countries, passports are issued to a citizen based on the submitted photo which allows the applicant to provide a morphed face photo to conceal his identity during the application process. In this work, we propose a novel approach leveraging the transferable features from a pre-trained Deep Convolutional Neural Networks (D-CNN) to detect both digital and print-scanned morphed face image. Thus, the proposed approach is based on the feature level fusion of the first fully connected layers of two D-CNN (VGG19 and AlexNet) that are specifically fine-tuned using the morphed face image database. The proposed method is extensively evaluated on the newly constructed database with both digital and print-scanned morphed face images corresponding to bona fide and morphed data reflecting a real-life scenario. The obtained results consistently demonstrate improved detection performance of the proposed scheme over previously proposed methods on both the digital and the print-scanned morphed face image database.","PeriodicalId":6668,"journal":{"name":"2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","volume":"146 1","pages":"1822-1830"},"PeriodicalIF":0.0,"publicationDate":"2017-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76611081","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}