The Experts below are selected from a list of 2508 Experts worldwide ranked by ideXlab platform
Sheikh Yaser - One of the best experts on this subject based on the ideXlab platform.
-
Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors
2018Co-Authors: Dong Xuanyi, Yu Shoou-i, Weng Xinshuo, Wei Shih-en, Yi Yang, Sheikh YaserAbstract:In this paper, we present supervision-by-registration, an unsupervised approach to improve the precision of facial Landmark detectors on both images and video. Our key observation is that the detections of the same Landmark in adjacent frames should be coherent with registration, i.e., optical flow. Interestingly, the coherency of optical flow is a source of supervision that does not require manual labeling, and can be leveraged during detector training. For example, we can enforce in the training loss function that a Detected Landmark at frame$_{t-1}$ followed by optical flow tracking from frame$_{t-1}$ to frame$_t$ should coincide with the location of the detection at frame$_t$. Essentially, supervision-by-registration augments the training loss function with a registration loss, thus training the detector to have output that is not only close to the annotations in labeled images, but also consistent with registration on large amounts of unlabeled videos. End-to-end training with the registration loss is made possible by a differentiable Lucas-Kanade operation, which computes optical flow registration in the forward pass, and back-propagates gradients that encourage temporal coherency in the detector. The output of our method is a more precise image-based facial Landmark detector, which can be applied to single images or video. With supervision-by-registration, we demonstrate (1) improvements in facial Landmark detection on both images (300W, ALFW) and video (300VW, Youtube-Celebrities), and (2) significant reduction of jittering in video detections.Comment: Minor modifications to the CVPR 2018 version (add missing references
Sheikh Y - One of the best experts on this subject based on the ideXlab platform.
-
Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors
'Institute of Electrical and Electronics Engineers (IEEE)', 2018Co-Authors: Dong X, Weng X, Se Wei, Yang Y, Sheikh YAbstract:© 2018 IEEE. In this paper, we present supervision-by-registration, an unsupervised approach to improve the precision of facial Landmark detectors on both images and video. Our key observation is that the detections of the same Landmark in adjacent frames should be coherent with registration, i.e., optical flow. Interestingly, coherency of optical flow is a source of supervision that does not require manual labeling, and can be leveraged during detector training. For example, we can enforce in the training loss function that a Detected Landmark at framet-1 followed by optical flow tracking from framet-1 to framet should coincide with the location of the detection at framet. Essentially, supervision-by-registration augments the training loss function with a registration loss, thus training the detector to have output that is not only close to the annotations in labeled images, but also consistent with registration on large amounts of unlabeled videos. End-to-end training with the registration loss is made possible by a differentiable Lucas-Kanade operation, which computes optical flow registration in the forward pass, and back-propagates gradients that encourage temporal coherency in the detector. The output of our method is a more precise image-based facial Landmark detector, which can be applied to single images or video. With supervision-by-registration, we demonstrate (1) improvements in facial Landmark detection on both images (300W, ALFW) and video (300VW, Youtube-Celebrities), and (2) significant reduction of jittering in video detections
Dong Xuanyi - One of the best experts on this subject based on the ideXlab platform.
-
Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors
2018Co-Authors: Dong Xuanyi, Yu Shoou-i, Weng Xinshuo, Wei Shih-en, Yi Yang, Sheikh YaserAbstract:In this paper, we present supervision-by-registration, an unsupervised approach to improve the precision of facial Landmark detectors on both images and video. Our key observation is that the detections of the same Landmark in adjacent frames should be coherent with registration, i.e., optical flow. Interestingly, the coherency of optical flow is a source of supervision that does not require manual labeling, and can be leveraged during detector training. For example, we can enforce in the training loss function that a Detected Landmark at frame$_{t-1}$ followed by optical flow tracking from frame$_{t-1}$ to frame$_t$ should coincide with the location of the detection at frame$_t$. Essentially, supervision-by-registration augments the training loss function with a registration loss, thus training the detector to have output that is not only close to the annotations in labeled images, but also consistent with registration on large amounts of unlabeled videos. End-to-end training with the registration loss is made possible by a differentiable Lucas-Kanade operation, which computes optical flow registration in the forward pass, and back-propagates gradients that encourage temporal coherency in the detector. The output of our method is a more precise image-based facial Landmark detector, which can be applied to single images or video. With supervision-by-registration, we demonstrate (1) improvements in facial Landmark detection on both images (300W, ALFW) and video (300VW, Youtube-Celebrities), and (2) significant reduction of jittering in video detections.Comment: Minor modifications to the CVPR 2018 version (add missing references
Dong X - One of the best experts on this subject based on the ideXlab platform.
-
Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors
'Institute of Electrical and Electronics Engineers (IEEE)', 2018Co-Authors: Dong X, Weng X, Se Wei, Yang Y, Sheikh YAbstract:© 2018 IEEE. In this paper, we present supervision-by-registration, an unsupervised approach to improve the precision of facial Landmark detectors on both images and video. Our key observation is that the detections of the same Landmark in adjacent frames should be coherent with registration, i.e., optical flow. Interestingly, coherency of optical flow is a source of supervision that does not require manual labeling, and can be leveraged during detector training. For example, we can enforce in the training loss function that a Detected Landmark at framet-1 followed by optical flow tracking from framet-1 to framet should coincide with the location of the detection at framet. Essentially, supervision-by-registration augments the training loss function with a registration loss, thus training the detector to have output that is not only close to the annotations in labeled images, but also consistent with registration on large amounts of unlabeled videos. End-to-end training with the registration loss is made possible by a differentiable Lucas-Kanade operation, which computes optical flow registration in the forward pass, and back-propagates gradients that encourage temporal coherency in the detector. The output of our method is a more precise image-based facial Landmark detector, which can be applied to single images or video. With supervision-by-registration, we demonstrate (1) improvements in facial Landmark detection on both images (300W, ALFW) and video (300VW, Youtube-Celebrities), and (2) significant reduction of jittering in video detections
Guoping Qiu - One of the best experts on this subject based on the ideXlab platform.
-
visual Landmark sequence based indoor localization
Proceedings of the 1st Workshop on Artificial Intelligence and Deep Learning for Geographic Knowledge Discovery, 2017Co-Authors: Jiasong Zhu, Tao Liu, Jon Garibaldi, Guoping QiuAbstract:This paper presents a method that uses common objects as Landmarks for smartphone-based indoor localization and navigation. First, a topological map marking relative positions of common objects such as doors, stairs and toilets is generated from floor plan. Second, a computer vision technique employing the latest deep learning technology has been developed for detecting common indoor objects from videos captured by smartphone. Third, second order Hidden Markov model is applied to match Detected indoor Landmark sequence to topological map. We use videos captured by users holding smartphones and walking through corridors of an office building to evaluate our method. The experiment shows that computer vision technique is able to accurately and reliably detect 10 classes of common indoor objects and that second order hidden Markov model can reliably match the Detected Landmark sequence with the topological map. This work demonstrates that computer vision and machine learning techniques can play a very useful role in developing smartphone-based indoor positioning applications.