Abstract
To solve deep metric learning problems and producing feature embeddings, current methodologies will commonly use a triplet model to minimise the relative distance between samples from the same class and maximise the relative distance between samples from different classes. Though successful, the training convergence of this triplet model can be compromised by the fact that the vast majority of the training samples will produce gradients with magnitudes that are close to zero. This issue has motivated the development of methods that explore the global structure of the embedding and other methods that explore hard negative/positive mining. The effectiveness of such mining methods is often associated with intractable computational requirements. In this paper, we propose a novel deep metric learning method that combines the triplet model and the global structure of the embedding space. We rely on a smart mining procedure that produces effective training samples for a low computational cost. In addition, we propose an adaptive controller that automatically adjusts the smart mining hyper-parameters and speeds up the convergence of the training process. We show empirically that our proposed method allows for fast and more accurate training of triplet ConvNets than other competing mining methods. Additionally, we show that our method achieves new state-of-the-art embedding results for CUB-200-2011 and Cars196 datasets.
| Original language | English |
|---|---|
| Title of host publication | Proceedings - 2017 IEEE International Conference on Computer Vision, ICCV 2017 |
| Editors | Rita Cucchiara, Yasuyuki Matsushita, Nicu Sebe, Stefano Soatto |
| Place of Publication | Piscataway NJ USA |
| Publisher | IEEE, Institute of Electrical and Electronics Engineers |
| Pages | 2840-2848 |
| Number of pages | 9 |
| ISBN (Electronic) | 9781538610329 |
| ISBN (Print) | 9781538610336 |
| DOIs | |
| Publication status | Published - 22 Dec 2017 |
| Event | IEEE International Conference on Computer Vision 2017 - Venice, Italy Duration: 22 Oct 2017 → 29 Oct 2017 Conference number: 16th http://iccv2017.thecvf.com/ |
Conference
| Conference | IEEE International Conference on Computer Vision 2017 |
|---|---|
| Abbreviated title | ICCV 2017 |
| Country | Italy |
| City | Venice |
| Period | 22/10/17 → 29/10/17 |
| Internet address |
Cite this
}
Smart Mining for Deep Metric Learning. / Harwood, Ben; Vijay Kumar, B. G.; Carneiro, Gustavo; Reid, Ian; Drummond, Tom.
Proceedings - 2017 IEEE International Conference on Computer Vision, ICCV 2017. ed. / Rita Cucchiara; Yasuyuki Matsushita; Nicu Sebe; Stefano Soatto. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2017. p. 2840-2848.Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Research › peer-review
TY - GEN
T1 - Smart Mining for Deep Metric Learning
AU - Harwood, Ben
AU - Vijay Kumar, B. G.
AU - Carneiro, Gustavo
AU - Reid, Ian
AU - Drummond, Tom
PY - 2017/12/22
Y1 - 2017/12/22
N2 - To solve deep metric learning problems and producing feature embeddings, current methodologies will commonly use a triplet model to minimise the relative distance between samples from the same class and maximise the relative distance between samples from different classes. Though successful, the training convergence of this triplet model can be compromised by the fact that the vast majority of the training samples will produce gradients with magnitudes that are close to zero. This issue has motivated the development of methods that explore the global structure of the embedding and other methods that explore hard negative/positive mining. The effectiveness of such mining methods is often associated with intractable computational requirements. In this paper, we propose a novel deep metric learning method that combines the triplet model and the global structure of the embedding space. We rely on a smart mining procedure that produces effective training samples for a low computational cost. In addition, we propose an adaptive controller that automatically adjusts the smart mining hyper-parameters and speeds up the convergence of the training process. We show empirically that our proposed method allows for fast and more accurate training of triplet ConvNets than other competing mining methods. Additionally, we show that our method achieves new state-of-the-art embedding results for CUB-200-2011 and Cars196 datasets.
AB - To solve deep metric learning problems and producing feature embeddings, current methodologies will commonly use a triplet model to minimise the relative distance between samples from the same class and maximise the relative distance between samples from different classes. Though successful, the training convergence of this triplet model can be compromised by the fact that the vast majority of the training samples will produce gradients with magnitudes that are close to zero. This issue has motivated the development of methods that explore the global structure of the embedding and other methods that explore hard negative/positive mining. The effectiveness of such mining methods is often associated with intractable computational requirements. In this paper, we propose a novel deep metric learning method that combines the triplet model and the global structure of the embedding space. We rely on a smart mining procedure that produces effective training samples for a low computational cost. In addition, we propose an adaptive controller that automatically adjusts the smart mining hyper-parameters and speeds up the convergence of the training process. We show empirically that our proposed method allows for fast and more accurate training of triplet ConvNets than other competing mining methods. Additionally, we show that our method achieves new state-of-the-art embedding results for CUB-200-2011 and Cars196 datasets.
UR - http://www.scopus.com/inward/record.url?scp=85041913153&partnerID=8YFLogxK
U2 - 10.1109/ICCV.2017.307
DO - 10.1109/ICCV.2017.307
M3 - Conference Paper
AN - SCOPUS:85041913153
SN - 9781538610336
SP - 2840
EP - 2848
BT - Proceedings - 2017 IEEE International Conference on Computer Vision, ICCV 2017
A2 - Cucchiara, Rita
A2 - Matsushita, Yasuyuki
A2 - Sebe, Nicu
A2 - Soatto, Stefano
PB - IEEE, Institute of Electrical and Electronics Engineers
CY - Piscataway NJ USA
ER -