On this page you can find the publication list of the Machine Learning and Perception Lab.
2022 |
|
![]() | Niki Martinel; Matteo Dunnhofer; Rita Pucci; Gian Luca Foresti; Christian Micheloni Lord of the Rings: Hanoi Pooling and Self-Knowledge Distillation for Fast and Accurate Vehicle Reidentification Journal Article IEEE Transactions on Industrial Informatics, 18 (1), pp. 87-96, 2022. @article{9387157, title = {Lord of the Rings: Hanoi Pooling and Self-Knowledge Distillation for Fast and Accurate Vehicle Reidentification}, author = {Niki Martinel and Matteo Dunnhofer and Rita Pucci and Gian Luca Foresti and Christian Micheloni}, doi = {10.1109/TII.2021.3068927}, year = {2022}, date = {2022-01-01}, journal = {IEEE Transactions on Industrial Informatics}, volume = {18}, number = {1}, pages = {87-96}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
2021 |
|
![]() | Matteo Dunnhofer; Antonino Furnari; Giovanni Maria Farinella; Christian Micheloni Is First Person Vision Challenging for Object Tracking? Inproceedings Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, pp. 2698-2710, 2021. @inproceedings{Dunnhofer_2021_ICCV, title = {Is First Person Vision Challenging for Object Tracking?}, author = {Matteo Dunnhofer and Antonino Furnari and Giovanni Maria Farinella and Christian Micheloni}, year = {2021}, date = {2021-10-01}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops}, pages = {2698-2710}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Asad Munir; Chengjin Lyu; Bart Goossens; Wilfried Philips; Christian Micheloni Resolution Based Feature Distillation for Cross Resolution Person Re-Identification Inproceedings Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, pp. 281-289, 2021. @inproceedings{Munir_2021_ICCV, title = {Resolution Based Feature Distillation for Cross Resolution Person Re-Identification}, author = {Asad Munir and Chengjin Lyu and Bart Goossens and Wilfried Philips and Christian Micheloni}, year = {2021}, date = {2021-10-01}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops}, pages = {281-289}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Rita Pucci; Christian Micheloni; Niki Martinel Self-Attention Agreement Among Capsules Inproceedings Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, pp. 272-280, 2021. @inproceedings{Pucci_2021_ICCV, title = {Self-Attention Agreement Among Capsules}, author = {Rita Pucci and Christian Micheloni and Niki Martinel}, year = {2021}, date = {2021-10-01}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops}, pages = {272-280}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Goutam Bhat; Martin Danelljan; Radu Timofte; ...; Rao Muhammad Umer; ...; Christian Micheloni NTIRE 2021 Challenge on Burst Super-Resolution: Methods and Results Inproceedings Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, pp. 613-626, 2021. @inproceedings{Bhat_2021_CVPR, title = {NTIRE 2021 Challenge on Burst Super-Resolution: Methods and Results}, author = {Goutam Bhat and Martin Danelljan and Radu Timofte and ... and Rao Muhammad Umer and ... and Christian Micheloni}, year = {2021}, date = {2021-06-01}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, pages = {613-626}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Rita Pucci; Christian Micheloni; Niki Martinel Collaborative Image and Object Level Features for Image Colourisation Inproceedings Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, pp. 2160-2169, 2021. @inproceedings{Pucci_2021_CVPR, title = {Collaborative Image and Object Level Features for Image Colourisation}, author = {Rita Pucci and Christian Micheloni and Niki Martinel}, year = {2021}, date = {2021-06-01}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, pages = {2160-2169}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Matteo Miani; Matteo Dunnhofer; Christian Micheloni; Andrea Marini; Nicola Baldo Surrogate Safety Measures Prediction at Multiple Timescales in V2P Conflicts Based on Gated Recurrent Unit Journal Article Sustainability, 13 (17), 2021, ISSN: 2071-1050. @article{su13179681, title = {Surrogate Safety Measures Prediction at Multiple Timescales in V2P Conflicts Based on Gated Recurrent Unit}, author = {Matteo Miani and Matteo Dunnhofer and Christian Micheloni and Andrea Marini and Nicola Baldo}, url = {https://www.mdpi.com/2071-1050/13/17/9681}, doi = {10.3390/su13179681}, issn = {2071-1050}, year = {2021}, date = {2021-01-01}, journal = {Sustainability}, volume = {13}, number = {17}, abstract = {Improving pedestrian safety at urban intersections requires intelligent systems that should not only understand the actual vehicle–pedestrian (V2P) interaction state but also proactively anticipate the event’s future severity pattern. This paper presents a Gated Recurrent Unit-based system that aims to predict, up to 3 s ahead in time, the severity level of V2P encounters, depending on the current scene representation drawn from on-board radars’ data. A car-driving simulator experiment has been designed to collect sequential mobility features on a cohort of 65 licensed university students who faced different V2P conflicts on a planned urban route. To accurately describe the pedestrian safety condition during the encounter process, a combination of surrogate safety indicators, namely TAdv (Time Advantage) and T2 (Nearness of the Encroachment), are considered for modeling. Due to the nature of these indicators, multiple recurrent neural networks are trained to separately predict T2 continuous values and TAdv categories. Afterwards, their predictions are exploited to label serious conflict interactions. As a comparison, an additional Gated Recurrent Unit (GRU) neural network is developed to directly predict the severity level of inner-city encounters. The latter neural model reaches the best performance on the test set, scoring a recall value of 0.899. Based on selected threshold values, the presented models can be used to label pedestrians near accident events and to enhance existing intelligent driving systems.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Improving pedestrian safety at urban intersections requires intelligent systems that should not only understand the actual vehicle–pedestrian (V2P) interaction state but also proactively anticipate the event’s future severity pattern. This paper presents a Gated Recurrent Unit-based system that aims to predict, up to 3 s ahead in time, the severity level of V2P encounters, depending on the current scene representation drawn from on-board radars’ data. A car-driving simulator experiment has been designed to collect sequential mobility features on a cohort of 65 licensed university students who faced different V2P conflicts on a planned urban route. To accurately describe the pedestrian safety condition during the encounter process, a combination of surrogate safety indicators, namely TAdv (Time Advantage) and T2 (Nearness of the Encroachment), are considered for modeling. Due to the nature of these indicators, multiple recurrent neural networks are trained to separately predict T2 continuous values and TAdv categories. Afterwards, their predictions are exploited to label serious conflict interactions. As a comparison, an additional Gated Recurrent Unit (GRU) neural network is developed to directly predict the severity level of inner-city encounters. The latter neural model reaches the best performance on the test set, scoring a recall value of 0.899. Based on selected threshold values, the presented models can be used to label pedestrians near accident events and to enhance existing intelligent driving systems. |
![]() | Rao M Umer; G L Foresti; C Micheloni Deep Iterative Residual Convolutional Network for Single Image Super-Resolution Inproceedings 2020 25th International Conference on Pattern Recognition (ICPR), pp. 1852-1858, 2021. @inproceedings{9412159, title = {Deep Iterative Residual Convolutional Network for Single Image Super-Resolution}, author = {Rao M Umer and G L Foresti and C Micheloni}, doi = {10.1109/ICPR48806.2021.9412159}, year = {2021}, date = {2021-01-01}, booktitle = {2020 25th International Conference on Pattern Recognition (ICPR)}, pages = {1852-1858}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Asad Munir; Niki Martinel; Christian Micheloni Self and Channel Attention Network for Person Re-Identification Inproceedings 2020 25th International Conference on Pattern Recognition (ICPR), pp. 4025-4031, 2021. @inproceedings{9413159, title = {Self and Channel Attention Network for Person Re-Identification}, author = {Asad Munir and Niki Martinel and Christian Micheloni}, doi = {10.1109/ICPR48806.2021.9413159}, year = {2021}, date = {2021-01-01}, booktitle = {2020 25th International Conference on Pattern Recognition (ICPR)}, pages = {4025-4031}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Rita Pucci; Christian Micheloni; Gian Luca Foresti; Niki Martinel Fixed simplex coordinates for angular margin loss in CapsNet Inproceedings 2020 25th International Conference on Pattern Recognition (ICPR), pp. 3042-3049, 2021. @inproceedings{9412241, title = {Fixed simplex coordinates for angular margin loss in CapsNet}, author = {Rita Pucci and Christian Micheloni and Gian Luca Foresti and Niki Martinel}, doi = {10.1109/ICPR48806.2021.9412241}, year = {2021}, date = {2021-01-01}, booktitle = {2020 25th International Conference on Pattern Recognition (ICPR)}, pages = {3042-3049}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Matteo Dunnhofer; Niki Martinel; Christian Micheloni Improving MRI-based Knee Disorder Diagnosis with Pyramidal Feature Details Inproceedings Medical Imaging with Deep Learning, 2021. @inproceedings{<LineBreak>dunnhofer2021improving, title = {Improving MRI-based Knee Disorder Diagnosis with Pyramidal Feature Details}, author = {Matteo Dunnhofer and Niki Martinel and Christian Micheloni}, url = {https://openreview.net/forum?id=7psPmlNffvg}, year = {2021}, date = {2021-01-01}, booktitle = {Medical Imaging with Deep Learning}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } | |
Chengjin Lyu; Patrick Heyer; Asad Munir; Ljiljana Platisa; Christian Micheloni; Bart Goossens; Wilfried Philips Visible-Thermal Pedestrian Detection via Unsupervised Transfer Learning Inproceedings 2021 the 5th International Conference on Innovation in Artificial Intelligence, pp. 158–163, Association for Computing Machinery, Xia men, China, 2021, ISBN: 9781450388634. @inproceedings{10.1145/3461353.3461369, title = {Visible-Thermal Pedestrian Detection via Unsupervised Transfer Learning}, author = {Chengjin Lyu and Patrick Heyer and Asad Munir and Ljiljana Platisa and Christian Micheloni and Bart Goossens and Wilfried Philips}, url = {https://doi.org/10.1145/3461353.3461369}, doi = {10.1145/3461353.3461369}, isbn = {9781450388634}, year = {2021}, date = {2021-01-01}, booktitle = {2021 the 5th International Conference on Innovation in Artificial Intelligence}, pages = {158–163}, publisher = {Association for Computing Machinery}, address = {Xia men, China}, series = {ICIAI 2021}, abstract = {Recently, pedestrian detection using visible-thermal pairs plays a key role in around-the-clock applications, such as public surveillance and autonomous driving. However, the performance of a well-trained pedestrian detector may drop significantly when it is applied to a new scenario. Normally, to achieve a good performance on the new scenario, manual annotation of the dataset is necessary, while it is costly and unscalable. In this work, an unsupervised transfer learning framework is proposed for visible-thermal pedestrian detection tasks. Given well-trained detectors from a source dataset, the proposed framework utilizes an iterative process to generate and fuse training labels automatically, with the help of two auxiliary single-modality detectors (visible and thermal). To achieve label fusion, the knowledge of daytime and nighttime is adopted to assign priorities to labels according to their illumination, which improves the quality of generated training labels. After each iteration, the existing detectors are updated using new training labels. Experimental results demonstrate that the proposed method obtains state-of-the-art performance without any manual training labels on the target dataset.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Recently, pedestrian detection using visible-thermal pairs plays a key role in around-the-clock applications, such as public surveillance and autonomous driving. However, the performance of a well-trained pedestrian detector may drop significantly when it is applied to a new scenario. Normally, to achieve a good performance on the new scenario, manual annotation of the dataset is necessary, while it is costly and unscalable. In this work, an unsupervised transfer learning framework is proposed for visible-thermal pedestrian detection tasks. Given well-trained detectors from a source dataset, the proposed framework utilizes an iterative process to generate and fuse training labels automatically, with the help of two auxiliary single-modality detectors (visible and thermal). To achieve label fusion, the knowledge of daytime and nighttime is adopted to assign priorities to labels according to their illumination, which improves the quality of generated training labels. After each iteration, the existing detectors are updated using new training labels. Experimental results demonstrate that the proposed method obtains state-of-the-art performance without any manual training labels on the target dataset. | |
![]() | Matteo Dunnhofer; Niki Martinel; Christian Micheloni Weakly-Supervised Domain Adaptation of Deep Regression Trackers via Reinforced Knowledge Distillation Journal Article IEEE Robotics and Automation Letters, 6 (3), pp. 5016-5023, 2021. @article{9394708, title = {Weakly-Supervised Domain Adaptation of Deep Regression Trackers via Reinforced Knowledge Distillation}, author = {Matteo Dunnhofer and Niki Martinel and Christian Micheloni}, doi = {10.1109/LRA.2021.3070816}, year = {2021}, date = {2021-01-01}, journal = {IEEE Robotics and Automation Letters}, volume = {6}, number = {3}, pages = {5016-5023}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
![]() | Wai Lun Leong; Niki Martinel; Sunan Huang; Christian Micheloni; Gian Luca Foresti; Rodney Swee Huat Teo An Intelligent Auto-Organizing Aerial Robotic Sensor Network System for Urban Surveillance Journal Article Journal of Intelligent & Robotic Systems, 102 (2), pp. 1–22, 2021. @article{leong2021intelligent, title = {An Intelligent Auto-Organizing Aerial Robotic Sensor Network System for Urban Surveillance}, author = {Wai Lun Leong and Niki Martinel and Sunan Huang and Christian Micheloni and Gian Luca Foresti and Rodney Swee Huat Teo}, year = {2021}, date = {2021-01-01}, journal = {Journal of Intelligent & Robotic Systems}, volume = {102}, number = {2}, pages = {1--22}, publisher = {Springer}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
2020 |
|
![]() | Matteo Dunnhofer; Niki Martinel; Christian Micheloni An Exploration of Target-Conditioned Segmentation Methods for Visual Object Trackers Inproceedings European Conference on Computer Vision (ECCV) Workshops, 2020. @inproceedings{Dunnhofer_2020_ECCV, title = {An Exploration of Target-Conditioned Segmentation Methods for Visual Object Trackers}, author = {Matteo Dunnhofer; Niki Martinel; Christian Micheloni}, year = {2020}, date = {2020-08-28}, booktitle = {European Conference on Computer Vision (ECCV) Workshops}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Rao Muhammad Umer; Christian Micheloni Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution Inproceedings European Conference on Computer Vision (ECCV) Workshops, 2020. @inproceedings{Umer_2020_ECCV, title = {Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution}, author = {Rao Muhammad Umer; Christian Micheloni}, year = {2020}, date = {2020-08-28}, booktitle = {European Conference on Computer Vision (ECCV) Workshops}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Matej Kristan; Aleš Leonardis; Jiří Matas; Michael Felsberg; Roman Pflugfelder; Joni-Kristian Kämäräinen; Martin Danelljan; Luka Čehovin Zajc; Alan Lukežič; ...; Christian Micheloni; ...; Gian Luca Foresti; ...; Matteo Dunnhofer The Eighth Visual Object Tracking VOT2020 Challenge Results Inproceedings European Conference on Computer Vision (ECCV) 2020 Workshops, 2020. @inproceedings{, title = {The Eighth Visual Object Tracking VOT2020 Challenge Results}, author = {Matej Kristan and Aleš Leonardis and Ji{ř}í Matas and Michael Felsberg and Roman Pflugfelder and Joni-Kristian Kämäräinen and Martin Danelljan and Luka {Č}ehovin Zajc and Alan Luke{ž}i{č} and ... and Christian Micheloni and ... and Gian Luca Foresti and ... and Matteo Dunnhofer }, year = {2020}, date = {2020-08-28}, booktitle = {European Conference on Computer Vision (ECCV) 2020 Workshops}, abstract = {The Visual Object Tracking challenge VOT2020 is the eighth annual tracker benchmarking activity organized by the VOT initiative. Results of 58 trackers are presented; many are state-of-the-art trackers published at major computer vision conferences or in journals in the recent years. The VOT2020 challenge was composed of five sub-challenges focusing on different tracking domains: (i) VOT-ST2020 challenge focused on short-term tracking in RGB, (ii) VOT-RT2020 challenge focused on ``real-time'' short-term tracking in RGB, (iii) VOT-LT2020 focused on long-term tracking namely coping with target disappearance and reappearance, (iv) VOT-RGBT2020 challenge focused on short-term tracking in RGB and thermal imagery and (v) VOT-RGBD2020 challenge focused on long-term tracking in RGB and depth imagery. Only the VOT-ST2020 datasets were refreshed. A significant novelty is introduction of a new VOT short-term tracking evaluation methodology, and introduction of segmentation ground truth in the VOT-ST2020 challenge -- bounding boxes will no longer be used in the VOT-ST challenges. A new VOT Python toolkit that implements all these novelites was introduced. Performance of the tested trackers typically by far exceeds standard baselines. The source code for most of the trackers is publicly available from the VOT page. The dataset, the evaluation kit and the results are publicly available at the challenge website (http://votchallenge.net).}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } The Visual Object Tracking challenge VOT2020 is the eighth annual tracker benchmarking activity organized by the VOT initiative. Results of 58 trackers are presented; many are state-of-the-art trackers published at major computer vision conferences or in journals in the recent years. The VOT2020 challenge was composed of five sub-challenges focusing on different tracking domains: (i) VOT-ST2020 challenge focused on short-term tracking in RGB, (ii) VOT-RT2020 challenge focused on ``real-time'' short-term tracking in RGB, (iii) VOT-LT2020 focused on long-term tracking namely coping with target disappearance and reappearance, (iv) VOT-RGBT2020 challenge focused on short-term tracking in RGB and thermal imagery and (v) VOT-RGBD2020 challenge focused on long-term tracking in RGB and depth imagery. Only the VOT-ST2020 datasets were refreshed. A significant novelty is introduction of a new VOT short-term tracking evaluation methodology, and introduction of segmentation ground truth in the VOT-ST2020 challenge -- bounding boxes will no longer be used in the VOT-ST challenges. A new VOT Python toolkit that implements all these novelites was introduced. Performance of the tested trackers typically by far exceeds standard baselines. The source code for most of the trackers is publicly available from the VOT page. The dataset, the evaluation kit and the results are publicly available at the challenge website (http://votchallenge.net). |
![]() | Rita Pucci; Christian Micheloni; Gian Luca Foresti; Niki Martinel Deep interactive encoding with capsule networks for image classification Journal Article Multimedia Tools and Applications, 2020. @article{Pucci2020, title = {Deep interactive encoding with capsule networks for image classification}, author = {Rita Pucci; Christian Micheloni; Gian Luca Foresti; Niki Martinel }, doi = {https://doi.org/10.1007/s11042-020-09455-8}, year = {2020}, date = {2020-08-26}, journal = {Multimedia Tools and Applications}, abstract = {With new architectures providing astonishing performance on many vision tasks, the interest in Convolutional Neural Networks (CNNs) has grown exponentially in the recent past. Such architectures, however, are not problem-free. For instance, one of the many issues is that they require a huge amount of labeled data and are not able to encode pose and deformation information. Capsule Networks (CapsNets) have been recently proposed as a solution to the issues related to CNNs. CapsNet achieved interesting results in images recognition by addressing pose and deformation encoding challenges. Despite their success, CapsNets are still an under-investigated architecture with respect to the more classical CNNs. Following the ideas of CapsNet, we propose to introduce Residual Capsule Network (ResNetCaps) and Dense Capsule Network (DenseNetCaps) to tackle the image recognition problem. With these two architectures, we expand the encoding phase of CapsNet by adding residual convolutional and densely connected convolutional blocks. In addition to this, we investigate the application of feature interaction methods between capsules to promote their cooperation while dealing with complex data. Experiments on four benchmark datasets demonstrate that the proposed approach performs better than existing solutions.}, keywords = {}, pubstate = {published}, tppubtype = {article} } With new architectures providing astonishing performance on many vision tasks, the interest in Convolutional Neural Networks (CNNs) has grown exponentially in the recent past. Such architectures, however, are not problem-free. For instance, one of the many issues is that they require a huge amount of labeled data and are not able to encode pose and deformation information. Capsule Networks (CapsNets) have been recently proposed as a solution to the issues related to CNNs. CapsNet achieved interesting results in images recognition by addressing pose and deformation encoding challenges. Despite their success, CapsNets are still an under-investigated architecture with respect to the more classical CNNs. Following the ideas of CapsNet, we propose to introduce Residual Capsule Network (ResNetCaps) and Dense Capsule Network (DenseNetCaps) to tackle the image recognition problem. With these two architectures, we expand the encoding phase of CapsNet by adding residual convolutional and densely connected convolutional blocks. In addition to this, we investigate the application of feature interaction methods between capsules to promote their cooperation while dealing with complex data. Experiments on four benchmark datasets demonstrate that the proposed approach performs better than existing solutions. |
![]() | Rao Muhammad Umer; Gian Luca Foresti; Christian Micheloni Deep Generative Adversarial Residual Convolutional Networks for Real-World Super-Resolution Inproceedings The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020. @inproceedings{Umer_2020_CVPR_Workshops, title = {Deep Generative Adversarial Residual Convolutional Networks for Real-World Super-Resolution}, author = {Rao Muhammad Umer and Gian Luca Foresti and Christian Micheloni}, url = {http://openaccess.thecvf.com/content_CVPRW_2020/html/w31/Umer_Deep_Generative_Adversarial_Residual_Convolutional_Networks_for_Real-World_Super-Resolution_CVPRW_2020_paper.html}, year = {2020}, date = {2020-06-01}, booktitle = {The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Andreas Lugmayr; Martin Danelljan; Radu Timofte NTIRE 2020 Challenge on Real-World Image Super-Resolution: Methods and Results Inproceedings Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020. @inproceedings{Lugmayr_2020_CVPR_Workshops, title = {NTIRE 2020 Challenge on Real-World Image Super-Resolution: Methods and Results}, author = {Andreas Lugmayr and Martin Danelljan and Radu Timofte}, year = {2020}, date = {2020-06-01}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Matteo Dunnhofer; Maria Antico; Fumio Sasazawa; Yu Takeda; Saskia Camps; Niki Martinel; Christian Micheloni; Gustavo Carneiro; Davide Fontanarosa Siam-U-Net: encoder-decoder siamese network for knee cartilage tracking in ultrasound images Journal Article Medical Image Analysis, 60 , pp. 101631, 2020, ISSN: 13618415. @article{Dunnhofer2020, title = {Siam-U-Net: encoder-decoder siamese network for knee cartilage tracking in ultrasound images}, author = {Matteo Dunnhofer and Maria Antico and Fumio Sasazawa and Yu Takeda and Saskia Camps and Niki Martinel and Christian Micheloni and Gustavo Carneiro and Davide Fontanarosa}, url = {https://linkinghub.elsevier.com/retrieve/pii/S1361841519301677}, doi = {10.1016/j.media.2019.101631}, issn = {13618415}, year = {2020}, date = {2020-02-01}, journal = {Medical Image Analysis}, volume = {60}, pages = {101631}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
![]() | N Martinel; G L Foresti; C Micheloni Deep Pyramidal Pooling With Attention for Person Re-Identification Journal Article IEEE Transactions on Image Processing, 29 , pp. 7306-7316, 2020. @article{9117031, title = {Deep Pyramidal Pooling With Attention for Person Re-Identification}, author = {N Martinel and G L Foresti and C Micheloni}, year = {2020}, date = {2020-01-01}, journal = {IEEE Transactions on Image Processing}, volume = {29}, pages = {7306-7316}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
![]() | Shadi Abpeikar; Mehdi Ghatee; Gian Luca Foresti; Christian Micheloni Adaptive neural tree exploiting expert nodes to classify high-dimensional data Journal Article Neural Networks, 124 , pp. 20 - 38, 2020, ISSN: 0893-6080. @article{ABPEIKAR202020, title = {Adaptive neural tree exploiting expert nodes to classify high-dimensional data}, author = {Shadi Abpeikar and Mehdi Ghatee and Gian Luca Foresti and Christian Micheloni}, url = {http://www.sciencedirect.com/science/article/pii/S0893608019304319}, doi = {https://doi.org/10.1016/j.neunet.2019.12.029}, issn = {0893-6080}, year = {2020}, date = {2020-01-01}, journal = {Neural Networks}, volume = {124}, pages = {20 - 38}, abstract = {Classification of high dimensional data suffers from curse of dimensionality and over-fitting. Neural tree is a powerful method which combines a local feature selection and recursive partitioning to solve these problems, but it leads to high depth trees in classifying high dimensional data. On the other hand, if less depth trees are used, the classification accuracy decreases or over-fitting increases. This paper introduces a novel Neural Tree exploiting Expert Nodes (NTEN) to classify high-dimensional data. It is based on a decision tree structure, whose internal nodes are expert nodes performing multi-dimensional splitting. Any expert node has three decision-making abilities. Firstly, they can select the most eligible neural network with respect to the data complexity. Secondly, they evaluate the over-fitting. Thirdly, they can cluster the features to jointly minimize redundancy and overlapping. To this aim, metaheuristic optimization algorithms including GA, NSGA-II, PSO and ACO are applied. Based on these concepts, any expert node splits a class when the over-fitting is low, and clusters the features when the over-fitting is high. Some theoretical results on NTEN are derived, and experiments on 35 standard data show that NTEN reaches good classification results, reduces tree depth without over-fitting and degrading accuracy.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Classification of high dimensional data suffers from curse of dimensionality and over-fitting. Neural tree is a powerful method which combines a local feature selection and recursive partitioning to solve these problems, but it leads to high depth trees in classifying high dimensional data. On the other hand, if less depth trees are used, the classification accuracy decreases or over-fitting increases. This paper introduces a novel Neural Tree exploiting Expert Nodes (NTEN) to classify high-dimensional data. It is based on a decision tree structure, whose internal nodes are expert nodes performing multi-dimensional splitting. Any expert node has three decision-making abilities. Firstly, they can select the most eligible neural network with respect to the data complexity. Secondly, they evaluate the over-fitting. Thirdly, they can cluster the features to jointly minimize redundancy and overlapping. To this aim, metaheuristic optimization algorithms including GA, NSGA-II, PSO and ACO are applied. Based on these concepts, any expert node splits a class when the over-fitting is low, and clusters the features when the over-fitting is high. Some theoretical results on NTEN are derived, and experiments on 35 standard data show that NTEN reaches good classification results, reduces tree depth without over-fitting and degrading accuracy. |
W L Leong; N Martinel; S Huang; C Micheloni; G L Foresti; R Teo Integrated Perception and Tactical Behaviours in an Auto-Organizing Aerial Sensor Network Inproceedings 2020 International Conference on Unmanned Aircraft Systems (ICUAS), pp. 429-438, 2020. @inproceedings{9214052, title = {Integrated Perception and Tactical Behaviours in an Auto-Organizing Aerial Sensor Network}, author = {W L Leong and N Martinel and S Huang and C Micheloni and G L Foresti and R Teo}, doi = {10.1109/ICUAS48674.2020.9214052}, year = {2020}, date = {2020-01-01}, booktitle = {2020 International Conference on Unmanned Aircraft Systems (ICUAS)}, pages = {429-438}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } | |
Asad Munir; Christian Micheloni Self Attention based multi branch Network for Person Re-Identification Inproceedings 2020 5th International Conference on Smart and Sustainable Technologies (SpliTech), pp. 1-5, 2020. @inproceedings{9243741, title = {Self Attention based multi branch Network for Person Re-Identification}, author = {Asad Munir and Christian Micheloni}, doi = {10.23919/SpliTech49282.2020.9243741}, year = {2020}, date = {2020-01-01}, booktitle = {2020 5th International Conference on Smart and Sustainable Technologies (SpliTech)}, pages = {1-5}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } | |
![]() | Asad Munir; Niki Martinel; Christian Micheloni Multi Branch Siamese Network For Person Re-Identification Inproceedings 2020 IEEE International Conference on Image Processing (ICIP), pp. 2351-2355, 2020. @inproceedings{9191115, title = {Multi Branch Siamese Network For Person Re-Identification}, author = {Asad Munir and Niki Martinel and Christian Micheloni}, doi = {10.1109/ICIP40778.2020.9191115}, year = {2020}, date = {2020-01-01}, booktitle = {2020 IEEE International Conference on Image Processing (ICIP)}, pages = {2351-2355}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Rita Pucci; Christian Micheloni; Gian Luca Foresti; Niki Martinel Is It a Plausible Colour? UCapsNet for Image Colourisation Miscellaneous Self-Supervised Learning Workshop at NeurIPS 2020, 2020. @misc{pucci2020plausible, title = {Is It a Plausible Colour? UCapsNet for Image Colourisation}, author = {Rita Pucci and Christian Micheloni and Gian Luca Foresti and Niki Martinel}, year = {2020}, date = {2020-01-01}, howpublished = {Self-Supervised Learning Workshop at NeurIPS 2020}, keywords = {}, pubstate = {published}, tppubtype = {misc} } | |
![]() | Matteo Dunnhofer; Niki Martinel; Christian Micheloni Tracking-by-Trackers with a Distilled and Reinforced Model Inproceedings Asian Conference on Computer Vision (ACCV), pp. 631–650, 2020, ISBN: 978-3-030-69532-3. @inproceedings{10.1007/978-3-030-69532-3_38, title = {Tracking-by-Trackers with a Distilled and Reinforced Model}, author = {Matteo Dunnhofer and Niki Martinel and Christian Micheloni}, isbn = {978-3-030-69532-3}, year = {2020}, date = {2020-01-01}, booktitle = {Asian Conference on Computer Vision (ACCV)}, pages = {631--650}, abstract = {Visual object tracking was generally tackled by reasoning independently on fast processing algorithms, accurate online adaptation methods, and fusion of trackers. In this paper, we unify such goals by proposing a novel tracking methodology that takes advantage of other visual trackers, offline and online. A compact student model is trained via the marriage of knowledge distillation and reinforcement learning. The first allows to transfer and compress tracking knowledge of other trackers. The second enables the learning of evaluation measures which are then exploited online. After learning, the student can be ultimately used to build (i) a very fast single-shot tracker, (ii) a tracker with a simple and effective online adaptation mechanism, (iii) a tracker that performs fusion of other trackers. Extensive validation shows that the proposed algorithms compete with real-time state-of-the-art trackers.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Visual object tracking was generally tackled by reasoning independently on fast processing algorithms, accurate online adaptation methods, and fusion of trackers. In this paper, we unify such goals by proposing a novel tracking methodology that takes advantage of other visual trackers, offline and online. A compact student model is trained via the marriage of knowledge distillation and reinforcement learning. The first allows to transfer and compress tracking knowledge of other trackers. The second enables the learning of evaluation measures which are then exploited online. After learning, the student can be ultimately used to build (i) a very fast single-shot tracker, (ii) a tracker with a simple and effective online adaptation mechanism, (iii) a tracker that performs fusion of other trackers. Extensive validation shows that the proposed algorithms compete with real-time state-of-the-art trackers. |
2019 |
|
![]() | Matej Kristan; Jiri Matas; Ales Leonardis; Michael Felsberg; Roman Pflugfelder; Joni-Kristian Kamarainen; Luka Cehovin Zajc; Ondrej Drbohlav; Alan Lukezic; ...; Niki Martinel; ...; Christian Micheloni; ...; Matteo Dunnhofer; ... The Seventh Visual Object Tracking VOT2019 Challenge Results Inproceedings The IEEE International Conference on Computer Vision (ICCV) Workshops, 2019. @inproceedings{Kristan_2019_ICCV, title = {The Seventh Visual Object Tracking VOT2019 Challenge Results}, author = {Matej Kristan and Jiri Matas and Ales Leonardis and Michael Felsberg and Roman Pflugfelder and Joni-Kristian Kamarainen and Luka Cehovin Zajc and Ondrej Drbohlav and Alan Lukezic and ... and Niki Martinel and ... and Christian Micheloni and ... and Matteo Dunnhofer and ...}, url = {http://openaccess.thecvf.com/content_ICCVW_2019/html/VOT/Kristan_The_Seventh_Visual_Object_Tracking_VOT2019_Challenge_Results_ICCVW_2019_paper.html}, year = {2019}, date = {2019-10-01}, booktitle = {The IEEE International Conference on Computer Vision (ICCV) Workshops}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Matteo Dunnhofer; Niki Martinel; Gian Luca Foresti; Christian Micheloni Visual Tracking by Means of Deep Reinforcement Learning and an Expert Demonstrator Inproceedings The IEEE International Conference on Computer Vision (ICCV) Workshops, 2019. @inproceedings{Dunnhofer_2019_ICCV, title = {Visual Tracking by Means of Deep Reinforcement Learning and an Expert Demonstrator}, author = {Matteo Dunnhofer and Niki Martinel and Gian Luca Foresti and Christian Micheloni}, url = {http://openaccess.thecvf.com/content_ICCVW_2019/html/VOT/Dunnhofer_Visual_Tracking_by_Means_of_Deep_Reinforcement_Learning_and_an_ICCVW_2019_paper.html}, year = {2019}, date = {2019-10-01}, booktitle = {The IEEE International Conference on Computer Vision (ICCV) Workshops}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
![]() | Niki Martinel; Gian Luca Foresti; Christian Micheloni Aggregating Deep Pyramidal Representations for Person Re-Identification Inproceedings The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019. @inproceedings{Martinel_2019_CVPR_Workshops, title = {Aggregating Deep Pyramidal Representations for Person Re-Identification}, author = {Niki Martinel and Gian Luca Foresti and Christian Micheloni}, url = {http://openaccess.thecvf.com/content_CVPRW_2019/html/TRMTMCT/Martinel_Aggregating_Deep_Pyramidal_Representations_for_Person_Re-Identification_CVPRW_2019_paper.html}, year = {2019}, date = {2019-06-01}, booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Niki Martinel; Gian Luca Foresti; Christian Micheloni Distributed person re-identification through network-wise rank fusion consensus Journal Article Pattern Recognition Letters, 124 , pp. 63–73, 2019, ISSN: 01678655. @article{Martinel2019, title = {Distributed person re-identification through network-wise rank fusion consensus}, author = {Niki Martinel and Gian Luca Foresti and Christian Micheloni}, url = {https://doi.org/10.1016/j.patrec.2018.12.015 https://linkinghub.elsevier.com/retrieve/pii/S0167865518309267}, doi = {10.1016/j.patrec.2018.12.015}, issn = {01678655}, year = {2019}, date = {2019-06-01}, journal = {Pattern Recognition Letters}, volume = {124}, pages = {63--73}, publisher = {Elsevier B.V.}, abstract = {The problem of re-identify persons across single disjoint camera-pairs has received great attention from the community. Despite this, when the re-identification process has to be carried out on a wide camera network additional problems arise and deny the direct application of existing solutions. Thus, a different approach has to be considered. In particular, existing approaches have neglected the importance of the network topology (i.e., the configuration of the monitored area) in such a process. To try filling such a gap, we propose a distributed person re-identification framework which brings in the following contributions: (i) a weighted camera matching cost that measures the re-identification performance between cameras in the network; (ii) a derivation of the distance vector algorithm that yields to network topology learning and allows us to prioritize and limit the cameras inquired for the re-identification; (iii) a network consensus weighted rank fusion solution that allows us to perform the re-identification in a robust fashion. Results on four benchmark datasets show that the proposed approach brings to significant network-wise re-identification improvements.}, keywords = {}, pubstate = {published}, tppubtype = {article} } The problem of re-identify persons across single disjoint camera-pairs has received great attention from the community. Despite this, when the re-identification process has to be carried out on a wide camera network additional problems arise and deny the direct application of existing solutions. Thus, a different approach has to be considered. In particular, existing approaches have neglected the importance of the network topology (i.e., the configuration of the monitored area) in such a process. To try filling such a gap, we propose a distributed person re-identification framework which brings in the following contributions: (i) a weighted camera matching cost that measures the re-identification performance between cameras in the network; (ii) a derivation of the distance vector algorithm that yields to network topology learning and allows us to prioritize and limit the cameras inquired for the re-identification; (iii) a network consensus weighted rank fusion solution that allows us to perform the re-identification in a robust fashion. Results on four benchmark datasets show that the proposed approach brings to significant network-wise re-identification improvements. | |
Marco Zamprogno; Marco Passon; Niki Martinel; Giuseppe Serra; Christian Micheloni; Carlo Tasso; Gian Luca Foresti Video-Based Convolutional Attention for Person Re-Identification Inproceedings International Conference on Image Analysis and Processing, 2019. @inproceedings{Zamprogno2019, title = {Video-Based Convolutional Attention for Person Re-Identification}, author = {Marco Zamprogno and Marco Passon and Niki Martinel and Giuseppe Serra and Christian Micheloni and Carlo Tasso and Gian Luca Foresti}, url = {https://arxiv.org/abs/1910.04856}, year = {2019}, date = {2019-01-01}, booktitle = {International Conference on Image Analysis and Processing}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } | |
Rita Pucci; Christian Micheloni; Vito Roberto; Gian Luca Foresti; Niki Martinel An Exploration of the Interaction Between capsules with ResNetCaps models Inproceedings International Conference on Distributed Smart Cameras, pp. 1–6, ACM Press, Trento, Italy, 2019, ISBN: 9781450371896. @inproceedings{Pucci2019, title = {An Exploration of the Interaction Between capsules with ResNetCaps models}, author = {Rita Pucci and Christian Micheloni and Vito Roberto and Gian Luca Foresti and Niki Martinel}, url = {http://dl.acm.org/citation.cfm?doid=3349801.3349804}, doi = {10.1145/3349801.3349804}, isbn = {9781450371896}, year = {2019}, date = {2019-01-01}, booktitle = {International Conference on Distributed Smart Cameras}, pages = {1--6}, publisher = {ACM Press}, address = {Trento, Italy}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } | |
Enkhtogtokh Togootogtokh; Sunan Huang; Wai Lun Leong; Rodney Teo; Gian Luca Foresti; Christian Micheloni; Niki Maritnel An Efficient Artificial Intelligence Framework for UAV Systems Inproceedings International Conference on Ubi-Media Computing, 2019. @inproceedings{Togootogtokh2019, title = {An Efficient Artificial Intelligence Framework for UAV Systems}, author = {Enkhtogtokh Togootogtokh and Sunan Huang and Wai Lun Leong and Rodney Teo and Gian Luca Foresti and Christian Micheloni and Niki Maritnel}, year = {2019}, date = {2019-01-01}, booktitle = {International Conference on Ubi-Media Computing}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } | |
Giuseppe Lisanti; Niki Martinel; Christian Micheloni; Alberto Del Bimbo; Gian Luca Foresti From person to group re-identification via unsupervised transfer of sparse features Journal Article Image and Vision Computing, 83-84 , pp. 29–38, 2019, ISSN: 02628856. @article{Lisanti2019, title = {From person to group re-identification via unsupervised transfer of sparse features}, author = {Giuseppe Lisanti and Niki Martinel and Christian Micheloni and Alberto {Del Bimbo} and Gian {Luca Foresti}}, url = {https://doi.org/10.1016/j.imavis.2019.02.009}, doi = {10.1016/j.imavis.2019.02.009}, issn = {02628856}, year = {2019}, date = {2019-01-01}, journal = {Image and Vision Computing}, volume = {83-84}, pages = {29--38}, publisher = {Elsevier B.V.}, abstract = {The visual association of a person appearing in the field of view of different cameras is today well known as Person Re-Identification. Current approaches find a solution to such a problem by considering persons as individuals, hence avoiding the fact that frequently they form groups or move in crowds. In such cases, the information acquired by neighboring individuals can provide relevant visual context to boost the performance in re-identifying persons within the group. In light of enriched information, groups re-identification encompasses additional problems to the common person re-identification ones, such as severe occlusions and changes in the relative position of people within the group. In this paper, the single person re-identification knowledge is transferred by means of a sparse dictionary learning to group re-identification. First, patches extracted from single person images are used to learn a dictionary of sparse atoms. This is used to obtain a sparsity-driven residual group representation that is exploited to perform group re-identification. To evaluate the performance of the proposed approach, we considered the i-LIDS groups dataset that is the only group re-identification publicly available dataset. The benchmark datasets for single person re-identification evaluation do not include group information, hence we collected two additional datasets under challenging scenarios and used them to validate our solution.}, keywords = {}, pubstate = {published}, tppubtype = {article} } The visual association of a person appearing in the field of view of different cameras is today well known as Person Re-Identification. Current approaches find a solution to such a problem by considering persons as individuals, hence avoiding the fact that frequently they form groups or move in crowds. In such cases, the information acquired by neighboring individuals can provide relevant visual context to boost the performance in re-identifying persons within the group. In light of enriched information, groups re-identification encompasses additional problems to the common person re-identification ones, such as severe occlusions and changes in the relative position of people within the group. In this paper, the single person re-identification knowledge is transferred by means of a sparse dictionary learning to group re-identification. First, patches extracted from single person images are used to learn a dictionary of sparse atoms. This is used to obtain a sparsity-driven residual group representation that is exploited to perform group re-identification. To evaluate the performance of the proposed approach, we considered the i-LIDS groups dataset that is the only group re-identification publicly available dataset. The benchmark datasets for single person re-identification evaluation do not include group information, hence we collected two additional datasets under challenging scenarios and used them to validate our solution. | |
2018 |
|
Sunan Huang; Rodney Teo; William Leong; Niki Martinel; Gian Luca Foresti; Christian Micheloni Coverage Control of Multiple Unmanned Aerial Vehicles: A Short Review Journal Article Unmanned Systems, 6 (2), pp. 131––144, 2018. @article{Huang2018, title = {Coverage Control of Multiple Unmanned Aerial Vehicles: A Short Review}, author = {Sunan Huang and Rodney Teo and William Leong and Niki Martinel and Gian Luca Foresti and Christian Micheloni}, doi = {10.1142/S2301385018400046}, year = {2018}, date = {2018-01-01}, journal = {Unmanned Systems}, volume = {6}, number = {2}, pages = {131----144}, keywords = {}, pubstate = {published}, tppubtype = {article} } | |
Niki Martinel; Gian Luca Foresti; Christian Micheloni Unsupervised Hashing with Neural Trees for Image Retrieval and Person Re-Identification Inproceedings International Conference on Distributed Smart Cameras, 2018, ISBN: 9781450365116. @inproceedings{Martinel2018, title = {Unsupervised Hashing with Neural Trees for Image Retrieval and Person Re-Identification}, author = {Niki Martinel and Gian Luca Foresti and Christian Micheloni}, url = {https://dl.acm.org/doi/10.1145/3243394.3243687}, isbn = {9781450365116}, year = {2018}, date = {2018-01-01}, booktitle = {International Conference on Distributed Smart Cameras}, abstract = {Recent vision studies have shown that learning compact codes is of paramount importance to improve the massive data processing while significantly reducing storage footprints. This has recently yielded to a surge of effort in learning compact and robust hash functions for image retrieval tasks. The majority of the existing lit- erature has been devoted to the exploration of deep hash functions, typically under supervised scenarios. Unsupervised hashing meth- ods have been less attractive due to the their difficulty in achieving satisfactory performance for the same objective. In this work, we propose a simple yet effective unsupervised hashing framework, which exploits the powerful visual representation capabilities of deep architectures and combines this within a tree structure in- volving a multi-path scheme. The key advantage of the proposed method is the ability to bring in a divide-and-conquer approach to reduce the complexity of the classification problem at each node of the tree without the need of labeled data. To validate the proposed solution, experimental results on two benchmark datasets for image retrieval and person re-identification have been computed.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Recent vision studies have shown that learning compact codes is of paramount importance to improve the massive data processing while significantly reducing storage footprints. This has recently yielded to a surge of effort in learning compact and robust hash functions for image retrieval tasks. The majority of the existing lit- erature has been devoted to the exploration of deep hash functions, typically under supervised scenarios. Unsupervised hashing meth- ods have been less attractive due to the their difficulty in achieving satisfactory performance for the same objective. In this work, we propose a simple yet effective unsupervised hashing framework, which exploits the powerful visual representation capabilities of deep architectures and combines this within a tree structure in- volving a multi-path scheme. The key advantage of the proposed method is the ability to bring in a divide-and-conquer approach to reduce the complexity of the classification problem at each node of the tree without the need of labeled data. To validate the proposed solution, experimental results on two benchmark datasets for image retrieval and person re-identification have been computed. | |
Matteo Chini; Niki Martinel; Matteo Dunnhofer; Carlo Ceschia; Christian Micheloni Unsupervised Smoke Detection in Normally Smoking Environments Inproceedings International Conference on Distributed Smart Cameras, pp. 1–6, ACM Press, New York, New York, USA, 2018, ISBN: 9781450365116. @inproceedings{Chini2018, title = {Unsupervised Smoke Detection in Normally Smoking Environments}, author = {Matteo Chini and Niki Martinel and Matteo Dunnhofer and Carlo Ceschia and Christian Micheloni}, url = {http://dl.acm.org/citation.cfm?doid=3243394.3243699}, doi = {10.1145/3243394.3243699}, isbn = {9781450365116}, year = {2018}, date = {2018-01-01}, booktitle = {International Conference on Distributed Smart Cameras}, pages = {1--6}, publisher = {ACM Press}, address = {New York, New York, USA}, abstract = {The problem of smoke detection through visual analytics is an open challenging problem. The existing literature has addressed the problem by mainly working on the best feature representation and by exploiting supervised solutions which consider the prob- lem of smoke detection as a binary classification one. Differently from such works, we consider the possibility that in some contexts sensing smokes is a common situation, but want to detect when there are significative fluctuations within this normal situation. In light of such a consideration, we propose an unsupervised solu- tion that leverages on the concept of anomaly detection. Different visual representations have been used together with a temporal smoothing function reduce the effects of noisy measurement. Such temporally smoothed representations are then exploited to learn a robust ânormalityâ model by means of a One-Class Support Vector Machine. A real prototype has been developed and exploited to collect a new dataset which has been considered to evaluate the proposed solution.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } The problem of smoke detection through visual analytics is an open challenging problem. The existing literature has addressed the problem by mainly working on the best feature representation and by exploiting supervised solutions which consider the prob- lem of smoke detection as a binary classification one. Differently from such works, we consider the possibility that in some contexts sensing smokes is a common situation, but want to detect when there are significative fluctuations within this normal situation. In light of such a consideration, we propose an unsupervised solu- tion that leverages on the concept of anomaly detection. Different visual representations have been used together with a temporal smoothing function reduce the effects of noisy measurement. Such temporally smoothed representations are then exploited to learn a robust ânormalityâ model by means of a One-Class Support Vector Machine. A real prototype has been developed and exploited to collect a new dataset which has been considered to evaluate the proposed solution. | |
Danilo Avola; Luigi Cinque; Gian Luca Foresti; Niki Martinel; Daniele Pannone; Claudio Piciarelli A UAV Video Dataset for Mosaicking and Change Detection From Low-Altitude Flights Journal Article IEEE Transactions on Systems, Man, and Cybernetics: Systems, pp. 1–11, 2018, ISSN: 2168-2216. @article{Avola2018a, title = {A UAV Video Dataset for Mosaicking and Change Detection From Low-Altitude Flights}, author = {Danilo Avola and Luigi Cinque and Gian Luca Foresti and Niki Martinel and Daniele Pannone and Claudio Piciarelli}, url = {http://ieeexplore.ieee.org/document/8303666/}, doi = {10.1109/TSMC.2018.2804766}, issn = {2168-2216}, year = {2018}, date = {2018-01-01}, journal = {IEEE Transactions on Systems, Man, and Cybernetics: Systems}, pages = {1--11}, keywords = {}, pubstate = {published}, tppubtype = {article} } | |
Niki Martinel; Gian Luca Foresti; Christian Micheloni Wide-Slice Residual Networks for Food Recognition Inproceedings The IEEE Winter Conference on Applications of Computer Vision (WACV), 2018. @inproceedings{Martinel2016e, title = {Wide-Slice Residual Networks for Food Recognition}, author = {Niki Martinel and Gian Luca Foresti and Christian Micheloni}, url = {http://arxiv.org/abs/1612.06543}, year = {2018}, date = {2018-01-01}, booktitle = {The IEEE Winter Conference on Applications of Computer Vision (WACV)}, abstract = {Food diary applications represent a tantalizing market. Such applications, based on image food recognition, opened to new challenges for computer vision and pattern recognition algorithms. Recent works in the field are focusing either on hand-crafted representations or on learning these by exploiting deep neural networks. Despite the success of such a last family of works, these generally exploit off-the shelf deep architectures to classify food dishes. Thus, the architectures are not cast to the specific problem. We believe that better results can be obtained if the deep architecture is defined with respect to an analysis of the food composition. Following such an intuition, this work introduces a new deep scheme that is designed to handle the food structure. Specifically, inspired by the recent success of residual deep network, we exploit such a learning scheme and introduce a slice convolution block to capture the vertical food layers. Outputs of the deep residual blocks are combined with the sliced convolution to produce the classification score for specific food categories. To evaluate our proposed architecture we have conducted experimental results on three benchmark datasets. Results demonstrate that our solution shows better performance with respect to existing approaches (e.g., a top-1 accuracy of 90.27% on the Food-101 challenging dataset).}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Food diary applications represent a tantalizing market. Such applications, based on image food recognition, opened to new challenges for computer vision and pattern recognition algorithms. Recent works in the field are focusing either on hand-crafted representations or on learning these by exploiting deep neural networks. Despite the success of such a last family of works, these generally exploit off-the shelf deep architectures to classify food dishes. Thus, the architectures are not cast to the specific problem. We believe that better results can be obtained if the deep architecture is defined with respect to an analysis of the food composition. Following such an intuition, this work introduces a new deep scheme that is designed to handle the food structure. Specifically, inspired by the recent success of residual deep network, we exploit such a learning scheme and introduce a slice convolution block to capture the vertical food layers. Outputs of the deep residual blocks are combined with the sliced convolution to produce the classification score for specific food categories. To evaluate our proposed architecture we have conducted experimental results on three benchmark datasets. Results demonstrate that our solution shows better performance with respect to existing approaches (e.g., a top-1 accuracy of 90.27% on the Food-101 challenging dataset). | |
Niki Martinel Accelerated low-rank sparse metric learning for person re-identification Journal Article Pattern Recognition Letters, 112 , pp. 234–240, 2018, ISSN: 01678655. @article{Martinel2018a, title = {Accelerated low-rank sparse metric learning for person re-identification}, author = {Niki Martinel}, url = {https://doi.org/10.1016/j.patrec.2018.07.033}, doi = {10.1016/j.patrec.2018.07.033}, issn = {01678655}, year = {2018}, date = {2018-01-01}, journal = {Pattern Recognition Letters}, volume = {112}, pages = {234--240}, publisher = {Elsevier B.V.}, abstract = {Person re-identification is an open and challenging problem in computer vision. A surge of effort has been spent design the best feature representation, and to learn either the transformation of such features across cameras or an optimal matching metric. Metric learning solutions which are currently in vogue in the field generally require a dimensionality reduction pre-processing stage to handle the high-dimensionality of the adopted feature representation. Such an approach is suboptimal and a better solution can be achieved by combining such a step in the metric learning process. Towards this objective, a low-rank matrix which projects the high-dimensional vectors to a low-dimensional manifold with a discriminative Euclidean distance is introduced. The goal is achieved with a stochastic accelerated proximal gradient method. Experiments on two public benchmark datasets show that better performances than state-of-the-art methods are achieved.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Person re-identification is an open and challenging problem in computer vision. A surge of effort has been spent design the best feature representation, and to learn either the transformation of such features across cameras or an optimal matching metric. Metric learning solutions which are currently in vogue in the field generally require a dimensionality reduction pre-processing stage to handle the high-dimensionality of the adopted feature representation. Such an approach is suboptimal and a better solution can be achieved by combining such a step in the metric learning process. Towards this objective, a low-rank matrix which projects the high-dimensional vectors to a low-dimensional manifold with a discriminative Euclidean distance is introduced. The goal is achieved with a stochastic accelerated proximal gradient method. Experiments on two public benchmark datasets show that better performances than state-of-the-art methods are achieved. | |
2017 |
|
Niki Martinel; Gian Luca Foresti; Christian Micheloni Person Reidentification in a Distributed Camera Network Framework Journal Article IEEE Transactions on Cybernetics, 47 (11), pp. 3530–3541, 2017, ISSN: 2168-2267. @article{Martinel2016a, title = {Person Reidentification in a Distributed Camera Network Framework}, author = {Niki Martinel and Gian Luca Foresti and Christian Micheloni}, url = {http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7479516 http://ieeexplore.ieee.org/document/7479516/}, doi = {10.1109/TCYB.2016.2568264}, issn = {2168-2267}, year = {2017}, date = {2017-11-01}, journal = {IEEE Transactions on Cybernetics}, volume = {47}, number = {11}, pages = {3530--3541}, keywords = {}, pubstate = {published}, tppubtype = {article} } | |
Jorge Garcia; Niki Martinel; Alfredo Gardel; Ignacio Bravo; Gian Luca Foresti; Christian Micheloni Discriminant Context Information Analysis for Post-Ranking Person Re-Identification Journal Article IEEE Transactions on Image Processing, 26 (4), pp. 1650–1665, 2017, ISSN: 1057-7149. @article{Garcia2017, title = {Discriminant Context Information Analysis for Post-Ranking Person Re-Identification}, author = {Jorge Garcia and Niki Martinel and Alfredo Gardel and Ignacio Bravo and Gian Luca Foresti and Christian Micheloni}, url = {http://ieeexplore.ieee.org/document/7815412/}, doi = {10.1109/TIP.2017.2652725}, issn = {1057-7149}, year = {2017}, date = {2017-04-01}, journal = {IEEE Transactions on Image Processing}, volume = {26}, number = {4}, pages = {1650--1665}, keywords = {}, pubstate = {published}, tppubtype = {article} } | |
Niki Martinel; Claudio Piciarelli; Christian Micheloni An Ensemble Feature Method for Food Classification Journal Article Machine Graphics and Vision, 26 (1), pp. 13––39, 2017. @article{Martinel2017, title = {An Ensemble Feature Method for Food Classification}, author = {Niki Martinel and Claudio Piciarelli and Christian Micheloni}, url = {http://yadda.icm.edu.pl/yadda/element/bwmeta1.element.baztech-9bc56afd-1d03-420b-a0bb-492cfd82ab5c}, year = {2017}, date = {2017-01-01}, journal = {Machine Graphics and Vision}, volume = {26}, number = {1}, pages = {13----39}, keywords = {}, pubstate = {published}, tppubtype = {article} } | |
Niki Martinel; Matteo Dunnhofer; Gian Luca Foresti; Christian Micheloni Person Re-Identification via Unsupervised Transfer of Learned Visual Representations Inproceedings International Conference on Distributed Smart Cameras, pp. 1–6, Stanford, CA, USA, 2017, ISBN: 9781450354875. @inproceedings{Martinel2017b, title = {Person Re-Identification via Unsupervised Transfer of Learned Visual Representations}, author = {Niki Martinel and Matteo Dunnhofer and Gian Luca Foresti and Christian Micheloni}, url = {https://dl.acm.org/doi/10.1145/3131885.3131923}, isbn = {9781450354875}, year = {2017}, date = {2017-01-01}, booktitle = {International Conference on Distributed Smart Cameras}, pages = {1--6}, address = {Stanford, CA, USA}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } | |
Matteo Chini; Niki Martinel; Stefano Spagnul; Christian Micheloni Temporally Smoothed Anomaly Detection in Continuous Fluids Inproceedings International Conference on Distributed Smart Cameras, pp. 7–12, Stanford, CA, USA, 2017. @inproceedings{Chini2017, title = {Temporally Smoothed Anomaly Detection in Continuous Fluids}, author = {Matteo Chini and Niki Martinel and Stefano Spagnul and Christian Micheloni}, url = {https://dl.acm.org/doi/10.1145/3131885.3131924}, year = {2017}, date = {2017-01-01}, booktitle = {International Conference on Distributed Smart Cameras}, pages = {7--12}, address = {Stanford, CA, USA}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } | |
Giuseppe Lisanti; Niki Martinel; Alberto Del Bimbo; Gian Luca Foresti Group Re-Identification via Unsupervised Transfer of Sparse Features Encoding Inproceedings Proceedings of the IEEE International Conference on Computer Vision (ICCV), pp. 2449–2458, Venice, Italy, 2017. @inproceedings{Lisanti2017, title = {Group Re-Identification via Unsupervised Transfer of Sparse Features Encoding}, author = {Giuseppe Lisanti and Niki Martinel and Alberto {Del Bimbo} and Gian Luca Foresti}, url = {http://openaccess.thecvf.com/content_iccv_2017/html/Lisanti_Group_Re-Identification_via_ICCV_2017_paper.html}, year = {2017}, date = {2017-01-01}, booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV)}, pages = {2449--2458}, address = {Venice, Italy}, abstract = {Person re-identification is best known as the problem of associating a single person that is observed from one or more disjoint cameras. The existing literature has mainly addressed such an issue, neglecting the fact that people usually move in groups, like in crowded scenarios. We be- lieve that the additional information carried by neighboring individuals provides a relevant visual context that can be exploited to obtain a more robust match of single persons within the group. Despite this, re-identifying groups of peo- ple compound the common single person re-identification problems by introducing changes in the relative position of persons within the group and severe self-occlusions. In this paper, we propose a solution for group re-identification that grounds on transferring knowledge from single person re- identification to group re-identification by exploiting sparse dictionary learning. First, a dictionary of sparse atoms is learned using patches extracted from single person im- ages. Then, the learned dictionary is exploited to obtain a sparsity-driven residual group representation, which is fi- nally matched to perform the re-identification. Extensive experiments on the i-LIDS groups and two newly collected datasets show that the proposed solution outperforms state- of-the-art approaches.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Person re-identification is best known as the problem of associating a single person that is observed from one or more disjoint cameras. The existing literature has mainly addressed such an issue, neglecting the fact that people usually move in groups, like in crowded scenarios. We be- lieve that the additional information carried by neighboring individuals provides a relevant visual context that can be exploited to obtain a more robust match of single persons within the group. Despite this, re-identifying groups of peo- ple compound the common single person re-identification problems by introducing changes in the relative position of persons within the group and severe self-occlusions. In this paper, we propose a solution for group re-identification that grounds on transferring knowledge from single person re- identification to group re-identification by exploiting sparse dictionary learning. First, a dictionary of sparse atoms is learned using patches extracted from single person im- ages. Then, the learned dictionary is exploited to obtain a sparsity-driven residual group representation, which is fi- nally matched to perform the re-identification. Extensive experiments on the i-LIDS groups and two newly collected datasets show that the proposed solution outperforms state- of-the-art approaches. | |
2016 |
|
Jorge García; Niki Martinel; Alfredo Gardel; Ignacio Bravo; Gian Luca Foresti; Christian Micheloni Modeling feature distances by orientation driven classifiers for person re-identification Journal Article Journal of Visual Communication and Image Representation, 38 , pp. 115–129, 2016, ISSN: 10473203. @article{Garcia2016, title = {Modeling feature distances by orientation driven classifiers for person re-identification}, author = {Jorge García and Niki Martinel and Alfredo Gardel and Ignacio Bravo and Gian Luca Foresti and Christian Micheloni}, url = {http://linkinghub.elsevier.com/retrieve/pii/S1047320316000353}, doi = {10.1016/j.jvcir.2016.02.009}, issn = {10473203}, year = {2016}, date = {2016-07-01}, journal = {Journal of Visual Communication and Image Representation}, volume = {38}, pages = {115--129}, abstract = {To tackle the re-identification challenges existing methods propose to directly match image features or to learn the transformation of features that undergoes between two cameras. Other methods learn optimal similarity measures. However, the performance of all these methods are strongly dependent from the person pose and orientation. We focus on this aspect and introduce three main contributions to the field: (i) to propose a method to extract multiple frames of the same person with different orientations in order to capture the complete person appearance; (ii) to learn the pairwise feature dissimilarities space (PFDS) formed by the subspaces of similar and different image pair orientations; and (iii) within each subspace, a classifier is trained to capture the multi-modal inter-camera transformation of pairwise image dissimilarities and to discriminate between positive and negative pairs. The experiments show the superior performance of the proposed approach with respect to state-of-the-art methods using two publicly available benchmark datasets.}, keywords = {}, pubstate = {published}, tppubtype = {article} } To tackle the re-identification challenges existing methods propose to directly match image features or to learn the transformation of features that undergoes between two cameras. Other methods learn optimal similarity measures. However, the performance of all these methods are strongly dependent from the person pose and orientation. We focus on this aspect and introduce three main contributions to the field: (i) to propose a method to extract multiple frames of the same person with different orientations in order to capture the complete person appearance; (ii) to learn the pairwise feature dissimilarities space (PFDS) formed by the subspaces of similar and different image pair orientations; and (iii) within each subspace, a classifier is trained to capture the multi-modal inter-camera transformation of pairwise image dissimilarities and to discriminate between positive and negative pairs. The experiments show the superior performance of the proposed approach with respect to state-of-the-art methods using two publicly available benchmark datasets. | |
Niki Martinel; Claudio Piciarelli; Christian Micheloni A supervised extreme learning committee for food recognition Journal Article Computer Vision and Image Understanding, 148 , pp. 67–86, 2016, ISSN: 1090235X. @article{Martinel2016d, title = {A supervised extreme learning committee for food recognition}, author = {Niki Martinel and Claudio Piciarelli and Christian Micheloni}, doi = {10.1016/j.cviu.2016.01.012}, issn = {1090235X}, year = {2016}, date = {2016-01-01}, journal = {Computer Vision and Image Understanding}, volume = {148}, pages = {67--86}, abstract = {Food recognition is an emerging topic in computer vision. The problem is being addressed especially in health-oriented systems where it is used as a support for food diary applications. The goal is to improve current food diaries, where the users have to manually insert their daily food intake, with an automatic recognition of the food type, quantity and consequent calories intake estimation. In addition to the classical recognition challenges, the food recognition problem is characterized by the absence of a rigid structure of the food and by large intra-class variations. To tackle such challenges, a food recognition system based on a committee classification is proposed. The aim is to provide a system capable of automatically choosing the optimal features for food recognition out of the existing plethora of available ones (e.g., color, texture, etc.). Following this idea, each committee member, i.e., an Extreme Learning Machine, is trained to specialize on a single feature type. Then, a Structural Support Vector Machine is exploited to produce the final ranking of possible matches by filtering out the irrelevant features and thus merging only the relevant ones. Experimental results show that the proposed system outperforms state-of-the-art works on four publicly available benchmark datasets.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Food recognition is an emerging topic in computer vision. The problem is being addressed especially in health-oriented systems where it is used as a support for food diary applications. The goal is to improve current food diaries, where the users have to manually insert their daily food intake, with an automatic recognition of the food type, quantity and consequent calories intake estimation. In addition to the classical recognition challenges, the food recognition problem is characterized by the absence of a rigid structure of the food and by large intra-class variations. To tackle such challenges, a food recognition system based on a committee classification is proposed. The aim is to provide a system capable of automatically choosing the optimal features for food recognition out of the existing plethora of available ones (e.g., color, texture, etc.). Following this idea, each committee member, i.e., an Extreme Learning Machine, is trained to specialize on a single feature type. Then, a Structural Support Vector Machine is exploited to produce the final ranking of possible matches by filtering out the irrelevant features and thus merging only the relevant ones. Experimental results show that the proposed system outperforms state-of-the-art works on four publicly available benchmark datasets. |