Abstract
In recent years, deep Convolutional Neural Networks (CNNs) have broken all records in salient object detection. However, training such a deep model requires a large amount of manual annotations. Our goal is to overcome this limitation by automatically converting an existing deep contour detection model into a salient object detection model without using any manual salient object masks. For this purpose, we have created a deep network architecture, namely Contour-to-Saliency Network (C2S-Net), by grafting a new branch onto a well-trained contour detection network. Therefore, our C2S-Net has two branches for performing two different tasks: (1) predicting contours with the original contour branch, and (2) estimating per-pixel saliency score of each image with the newly-added saliency branch. To bridge the gap between these two tasks, we further propose a contour-to-saliency transferring method to automatically generate salient object masks which can be used to train the saliency branch from outputs of the contour branch. Finally, we introduce a novel alternating training pipeline to gradually update the network parameters. In this scheme, the contour branch generates saliency masks for training the saliency branch, while the saliency branch, in turn, feeds back saliency knowledge in the form of saliency-aware contour labels, for fine-tuning the contour branch. The proposed method achieves state-of-the-art performance on five well-known benchmarks, outperforming existing fully supervised methods while also maintaining high efficiency.
| Original language | English |
|---|---|
| Title of host publication | Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings |
| Editors | Yair Weiss, Vittorio Ferrari, Cristian Sminchisescu, Martial Hebert |
| Publisher | Springer Verlag |
| Pages | 370-385 |
| Number of pages | 16 |
| ISBN (Print) | 9783030012663 |
| DOIs | |
| Publication status | Published - 2018 Jan 1 |
| Externally published | Yes |
| Event | 15th European Conference on Computer Vision, ECCV 2018 - Munich, Germany Duration: 2018 Sep 8 → 2018 Sep 14 |
Publication series
| Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
|---|---|
| Volume | 11219 LNCS |
| ISSN (Print) | 0302-9743 |
| ISSN (Electronic) | 1611-3349 |
Other
| Other | 15th European Conference on Computer Vision, ECCV 2018 |
|---|---|
| Country | Germany |
| City | Munich |
| Period | 18/9/8 → 18/9/14 |
Fingerprint
Keywords
- Deep learning
- Saliency detection
- Transfer learning
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)
Cite this
Contour knowledge transfer for salient object detection. / Li, Xin; Yang, Fan; Cheng, Hong; Liu, Wei; Shen, Dinggang.
Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. ed. / Yair Weiss; Vittorio Ferrari; Cristian Sminchisescu; Martial Hebert. Springer Verlag, 2018. p. 370-385 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11219 LNCS).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Contour knowledge transfer for salient object detection
AU - Li, Xin
AU - Yang, Fan
AU - Cheng, Hong
AU - Liu, Wei
AU - Shen, Dinggang
PY - 2018/1/1
Y1 - 2018/1/1
N2 - In recent years, deep Convolutional Neural Networks (CNNs) have broken all records in salient object detection. However, training such a deep model requires a large amount of manual annotations. Our goal is to overcome this limitation by automatically converting an existing deep contour detection model into a salient object detection model without using any manual salient object masks. For this purpose, we have created a deep network architecture, namely Contour-to-Saliency Network (C2S-Net), by grafting a new branch onto a well-trained contour detection network. Therefore, our C2S-Net has two branches for performing two different tasks: (1) predicting contours with the original contour branch, and (2) estimating per-pixel saliency score of each image with the newly-added saliency branch. To bridge the gap between these two tasks, we further propose a contour-to-saliency transferring method to automatically generate salient object masks which can be used to train the saliency branch from outputs of the contour branch. Finally, we introduce a novel alternating training pipeline to gradually update the network parameters. In this scheme, the contour branch generates saliency masks for training the saliency branch, while the saliency branch, in turn, feeds back saliency knowledge in the form of saliency-aware contour labels, for fine-tuning the contour branch. The proposed method achieves state-of-the-art performance on five well-known benchmarks, outperforming existing fully supervised methods while also maintaining high efficiency.
AB - In recent years, deep Convolutional Neural Networks (CNNs) have broken all records in salient object detection. However, training such a deep model requires a large amount of manual annotations. Our goal is to overcome this limitation by automatically converting an existing deep contour detection model into a salient object detection model without using any manual salient object masks. For this purpose, we have created a deep network architecture, namely Contour-to-Saliency Network (C2S-Net), by grafting a new branch onto a well-trained contour detection network. Therefore, our C2S-Net has two branches for performing two different tasks: (1) predicting contours with the original contour branch, and (2) estimating per-pixel saliency score of each image with the newly-added saliency branch. To bridge the gap between these two tasks, we further propose a contour-to-saliency transferring method to automatically generate salient object masks which can be used to train the saliency branch from outputs of the contour branch. Finally, we introduce a novel alternating training pipeline to gradually update the network parameters. In this scheme, the contour branch generates saliency masks for training the saliency branch, while the saliency branch, in turn, feeds back saliency knowledge in the form of saliency-aware contour labels, for fine-tuning the contour branch. The proposed method achieves state-of-the-art performance on five well-known benchmarks, outperforming existing fully supervised methods while also maintaining high efficiency.
KW - Deep learning
KW - Saliency detection
KW - Transfer learning
UR - http://www.scopus.com/inward/record.url?scp=85055424049&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85055424049&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-01267-0_22
DO - 10.1007/978-3-030-01267-0_22
M3 - Conference contribution
SN - 9783030012663
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 370
EP - 385
BT - Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings
A2 - Weiss, Yair
A2 - Ferrari, Vittorio
A2 - Sminchisescu, Cristian
A2 - Hebert, Martial
PB - Springer Verlag
ER -