TY - GEN
T1 - Built-in foreground/background prior for weakly-supervised semantic segmentation
AU - Saleh, Fatemehsadat
AU - Akbarian, Mohammad Sadegh Ali
AU - Salzmann, Mathieu
AU - Petersson, Lars
AU - Gould, Stephen
AU - Alvarez, Jose M.
N1 - Publisher Copyright:
© Springer International Publishing AG 2016.
PY - 2016
Y1 - 2016
N2 - Pixel-level annotations are expensive and time consuming to obtain. Hence, weak supervision using only image tags could have a significant impact in semantic segmentation. Recently, CNN-based methods have proposed to fine-tune pre-trained networks using image tags. Without additional information, this leads to poor localization accuracy. This problem, however, was alleviated by making use of objectness priors to generate foreground/background masks. Unfortunately these priors either require training pixel-level annotations/bounding boxes, or still yield inaccurate object boundaries. Here, we propose a novel method to extract markedly more accurate masks from the pre-trained network itself, forgoing external objectness modules. This is accomplished using the activations of the higher-level convolutional layers, smoothed by a dense CRF. We demonstrate that our method, based on these masks and a weakly-supervised loss, outperforms the state-of-the-art tag-based weakly-supervised semantic segmentation techniques. Furthermore, we introduce a new form of inexpensive weak supervision yielding an additional accuracy boost.
AB - Pixel-level annotations are expensive and time consuming to obtain. Hence, weak supervision using only image tags could have a significant impact in semantic segmentation. Recently, CNN-based methods have proposed to fine-tune pre-trained networks using image tags. Without additional information, this leads to poor localization accuracy. This problem, however, was alleviated by making use of objectness priors to generate foreground/background masks. Unfortunately these priors either require training pixel-level annotations/bounding boxes, or still yield inaccurate object boundaries. Here, we propose a novel method to extract markedly more accurate masks from the pre-trained network itself, forgoing external objectness modules. This is accomplished using the activations of the higher-level convolutional layers, smoothed by a dense CRF. We demonstrate that our method, based on these masks and a weakly-supervised loss, outperforms the state-of-the-art tag-based weakly-supervised semantic segmentation techniques. Furthermore, we introduce a new form of inexpensive weak supervision yielding an additional accuracy boost.
KW - Convolutional neural networks
KW - Semantic segmentation
KW - Weak annotation
KW - Weakly-supervised segmentation
UR - http://www.scopus.com/inward/record.url?scp=84990045050&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-46484-8_25
DO - 10.1007/978-3-319-46484-8_25
M3 - Conference contribution
SN - 9783319464831
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 413
EP - 432
BT - Computer Vision - 14th European Conference, ECCV 2016, Proceedings
A2 - Leibe, Bastian
A2 - Matas, Jiri
A2 - Sebe, Nicu
A2 - Welling, Max
PB - Springer Verlag
T2 - 14th European Conference on Computer Vision, ECCV 2016
Y2 - 8 October 2016 through 16 October 2016
ER -